Welsh Assembly Government · Appendix C. Study characteristics of the primary studies 51 C.1...
Transcript of Welsh Assembly Government · Appendix C. Study characteristics of the primary studies 51 C.1...
Welsh Assembly Government
Outcomes-based Allocation of Educational
Resources
Rapid Evidence Assessment
Final report
22 December 2009
REA: Educational resources and outcomes
Matrix Evidence | 20 January 2011 2
Disclaimer In keeping with our values of integrity and excellence, Matrix has taken reasonable professional care in the preparation of this report. Although Matrix has made reasonable efforts to obtain information from a broad spectrum of sources, we cannot guarantee absolute accuracy or completeness of information/data submitted, nor do we accept responsibility for recommendations that may have been omitted due to particular or exceptional conditions and circumstances.
Confidentiality This report has been prepared for the client within the terms of our contract, and contains information which is proprietary to Matrix and confidential to our relationship. This may not be disclosed to third parties without prior agreement. Except where permitted under the provisions of confidentiality above, this document may not be reproduced, retained or stored beyond the period of validity, or transmitted in whole, or in part, without Matrix Evidence‘s prior, written permission. © Matrix Evidence Ltd, 2009 Any enquiries about this report should be directed to [email protected]
REA: Educational resources and outcomes
Matrix Evidence | 20 January 2011 3
Abstract
We conducted a rapid evidence assessment (REA) of the literature on the relation between
educational resourcing and various outcomes. Of nearly 4,000 abstracts assessed, 60 studies
were identified and synthesised using systematic REA methods. Overall, the evidence suggests
that there is a small positive association between educational inputs (such as per-pupil
expenditure) and attainment. There is some evidence for improvements in educational
progression (typically measured as pursuing higher education) and reduction in likely crime-
related behaviour, although fewer studies measured these outcomes. The study concludes that
there might be sufficient grounds for using attainment and educational progression in an
outcomes-based educational resource allocation formula, but a well-developed dataset with
pupil-level data would be required to establish the strength of the relation in the Welsh context.
Keywords
Economics of education
Educational resources
Educational expenditure
Student-teacher ratio
Class size
Educational grants
Educational outcomes
Attainment
REA: Educational resources and outcomes
Matrix Evidence | 20 January 2011 4
Contents
1.0 Key implications 5
2.0 Executive summary 6
3.0 Main report 9
3.1 Context 9
3.2 Approach 12 3.2.1 Research questions 13 3.2.2 Searching 13 3.2.3 Screening 14 3.2.4 Data extraction 15 3.2.5 Data synthesis 15 3.2.6 Summary 16
3.3 Results 16 3.3.1 Studies located 16 3.3.2 Summary of key study characteristics 17 3.3.3 Resources and student attainment 17 3.3.4 Resources and other outcomes 31 3.3.5 Summary of the findings 36
3.4 Implications 37
3.5 Knowledge gaps 39
3.6 References 42
Appendix A. Search Strategy, inclusion criteria, and inclusion
figures 46
A.1 Search sources 46
A.2 Search strategy 47
A.3 Screening 48
Appendix B. Flow of literature through the review 50
Appendix C. Study characteristics of the primary studies 51
C.1 Descriptive statistics 51
C.2 Limitations of the Research Field 11
C.3 Limitations with the individual studies included in this review (an annotated
bibliography) 52
Appendix D. Data extraction tool 58
Appendix E. Summary of included studies 60
REA: Educational resources and outcomes
Matrix Evidence | 20 January 2011 5
1.0 Key implications
In the context of considering options for developing and refining funding formulae for schools in
Wales to better reflect desirable educational outcomes, this study set out to review the evidence
for links between school funding and pupil attainment, as well as a range of other pupil
outcomes including university entrance, truancy, criminal behaviour, and future earnings. Key
policy implications arising from the review findings are summarised below:
There is evidence for a link between school funding and outcomes. However, it is most
meaningful to consider this association in relation to pupil progress (i.e., ‗value-added‘
attainment);
Evidence shows clearly that pupil attainment is also influenced by individual, social, and
economic factors over which schools have no control. For that reason, value-added
approaches to funding formulae are likely to be more equitable than approaches based
solely on pupil attainment;
Since links between funding and pupil attainment are much less consistent for primary
schools, it may be that different funding formulae would be appropriate for the primary
and secondary sectors;
Given that the evidence suggests the strength of links between funding and pupil
attainment varies across subject areas, consideration should be given as to the
measures of attainment included in an outcome-based funding formula;
The strength of evidence for links between funding and pupil attainment varies across
countries. The development of a funding formulae should be informed by reference to
data from Welsh schools. Following from this, it would be important to establish that the
relevant datasets are fit for purpose;
The rate of progression to higher education might be improved by increased resourcing,
according to the UK evidence;
Truancy appears to be reduced and attendance improved by increased resourcing,
according to a small number of studies;
The likelihood of future involvement in crime-related behaviours, such as juvenile
delinquency, appears to reduce with increased educational resourcing, according to a
small number of studies;
Future wages, occupational or educational aspirations, and social outcomes are not
consistently related to resourcing, according to the few studies reporting such outcomes
that were reviewed here.
REA: Educational resources and outcomes
Matrix Evidence | 20 January 2011 6
2.0 Executive summary
The review described in this report set out to establish the extent to which research evidence
supports the development of a new funding formula for schools in Wales based on various
outcomes. Specifically, the review examines the evidence for links between school funding and
pupil attainment, as well as a range of other pupil outcomes including university entrance,
truancy, criminal behaviour, and future earnings.
The review identified, retrieved, and reviewed studies from the academic literature using an
approach known as rapid evidence assessment (REA). REAs involve transparent and explicit
methods applied in a standardised and systematic way, producing a synthesis of the evidence
that is, as far as possible, accurate and free from bias.
Searching yielded nearly 4,000 studies of possible relevance. The abstracts and titles of these
studies were then assessed on their relevance to the review question, which narrowed the
sample down to 298 studies. The full-text version of these studies were then retrieved and
screened. The final result was a sample of 60 studies produced since 1999 that were deemed
to be both relevant and to have sufficient information to warrant synthesis.
The review team extracted data from these 60 studies including the country in which the study
was conducted, the educational input type and the findings in relation to pupil outcomes. They
then synthesised the extracted information to summarise the results of the studies and the
extent to which we can be confident in the findings. We present a summary of the findings
below.
Before summarising the results, it is important to understand the limitations of this review. Some
of the evidence we have reviewed has been extracted from studies low on methodological
rigour, which inevitably reduce confidence the robustness of their findings. Importantly, many of
the studies in the review are correlational in nature. Such research designs do not allow causal
attributions to be made. In other words, very few of the studies included here allow us to
definitively say that increased educational inputs cause improved outcomes.
More generally, it is important to understand that pupil attainment can is inexorably influenced
by a complex and wide-ranging set of factors, many of which fall outside of the school
environment. Relationships between funding and pupil outcomes will be subject to the influence
by any number of external or mediating factors. For example, Bramley and Watkins (2007)
noted that the best predictor of academic attainment is prior attainment; examining only the link
between funding and attainment without considering students‘ previous academic attainment
will lead to an overestimation of the contribution of funding to subsequent attainment. Numerous
other variables, such as the social and economic status of the student, or their belief in their
own academic ability, are also likely to play a role in determining future attainment (Marsh &
O‘Mara, 2008) and might also mediate the resource-attainment link.
REA: Educational resources and outcomes
Matrix Evidence | 20 January 2011 7
Summary of findings in relation to educational attainment
Fifty of the studies examined pupil attainment as an outcome. UK studies tend to support a
small, positive association between per-pupil expenditure and attainment. However, both UK
and international research suggests that the impact of funding on attainment may be subject-
dependent. In general, spending has a greater impact on attainment in maths and science than
on English/verbal attainment and general school grades. The evidence also suggests that
although funding increases might benefit secondary-school students, the evidence for primary-
school children is more mixed.
Findings on the impact of class size and attainment are mixed. UK research suggests that
improving the pupil-teacher ratios (PTR) has a positive impact on science and possibly maths
attainment, but not on English attainment. However, research from other countries has
generally not found a significant relationship between PTR and attainment.
Increasing funding for ICT and library facilities might improve attainment, although findings are
mixed. Grant initiatives, in which funding is dedicated to broad improvement programmes,
appear to be associated with small but positive effects on attainment.
More generally, the evidence around how funding might influence pupil attainment needs to be
interpreted with care. Learning and attainment take place in the context of complex social and
individual circumstances. In particular, evidence has shown that pupil attainment is influenced
by prior attainment, by the pupils‘ social and economic circumstances and by the quality of the
teaching that they receive.
Summary of findings in relation to other outcomes
Evidence concerning the impact of funding on other pupil outcomes is relatively rare, making it
harder to draw firm conclusions. Four UK studies on educational progression (e.g., continuing to
higher education) suggest that it might be improved by increased resourcing. However, four US
studies failed to find evidence linking improved financial resources to educational progression.
The small number of studies looking at truancy and attendance suggest increased resourcing
can lead to positive pupil outcomes. Similarly, a small number of studies suggest the risk of
committing crime-related behaviours (e.g., juvenile delinquency) might be reduced by increasing
resources—particularly where those resources are used to reduce the pupil-teacher ratio.
Evidence from three studies reviewed found no clear relation between school resourcing and
pupils‘ future earnings. Similarly, no conclusions can be drawn regarding the relation between
resourcing and pupils‘ aspirations or social outcomes.
How confident can we be that resource inputs cause outcomes?
Causality in research methods typically refers to the level of confidence in our findings that
differences in an outcome are caused by a particular input (e.g., ―increased expenditure causes
REA: Educational resources and outcomes
Matrix Evidence | 20 January 2011 8
improved pupil outcomes‖). Experimental methods, in which individuals are randomly assigned
to a ―treated‖ group or ―untreated‖ group, are the strongest designs for evaluating the causal
effect of an input on an outcome. Very few of the studies in this review employed experimental
research methods. Most of the studies examined associations between resourcing and
attainment over time (i.e., longitudinal modelling). Such methods make it difficult to claim
confidently any causal relationship between inputs and outcomes because longitudinal research
designs cannot fully take into account initial differences between individuals, or other variables
that might influence the relation between input and outcome. As such, the findings presented
herein generally do not verify a causal relation, but rather suggest that there may be an
association between these variables over time that appear to have some causal ordering (that
is, prior inputs can predict future outcomes).
Implications for developing an outcomes-based funding formula
The review has found sufficient evidence to support the view that levels of educational funding
can influence pupil attainment; it also found that funding might have an impact on educational
progression and crime-related behaviours, although the pool of evidence on these outcomes is
smaller. However, when determining the viability of a funding formula based on this evidence,
the Welsh Assembly Government needs to be aware that the relationship between funding and
attainment is inevitably influenced by pupil characteristics, staff quality and programmes
available at the school (such as Reading Recovery literacy programmes or special educational
needs units).
Key to predicting the impact of spending on attainment is considering how funding is spent once
it is obtained. The Welsh system is quite devolved, in that schools have a high degree of
autonomy over what they do with their funds. Should an outcomes-based formula be
introduced, it might be useful to offer guidance to schools on the types of initiatives that they
could implement to maximise the effectiveness of their additional resources. A review of the
literature on successful educational intervention might be useful in developing such advice.
Furthermore, although we have established that there appears to be some evidence of a
relation between resourcing and certain outcomes (attainment, educational progression and
crime-related behaviours), we have not established the strength of this relation. This could be
confirmed for Welsh students through the statistical analysis of large-scale databases (e.g.,
PLASC - Pupil Level Annual School Census) and possibly supplemented with quantitative meta-
analysis. In other words, to implement an outcomes based school funding model it would need
to be established how much additional funding per pupil with a given level of disadvantage is
needed in order to get their attainment to a defined level. Bramley and Watkins (2007) produced
a strong report that takes us closer to this goal1.
1 This report is very strong and uses appropriate multilevel model analyses. However, they acknowledge that the
unavailability of some data (e.g., school capacity and capital stock) could be a limitation of the analyses. As such, some updating of the analyses may be required when such data are available and adequate.
REA: Educational resources and outcomes
Matrix Evidence | 20 January 2011 9
3.0 Main report
3.1 Policy context
This report summarises the evidence produced since 1999 on the relationship between
educational resources for primary and secondary students and various outcomes2. The
objective of the study was to provide context for the possible future development of a new
outcomes-based resource allocation formula for the Welsh Assembly Government, by
establishing whether resources are sufficiently linked to outcomes. The report does not seek to
determine the how much additional spending is required to obtain a desired improvement in
student outcomes, which could be better achieved through sophisticated statistical modelling
such as in Bramley and Watkins (2007).
The Welsh Assembly Government provides funding to local authorities for pre-16 provision in
schools in Wales, mainly through the local government revenue settlement (Revenue Support
Grant). The other main sources of funding for local authority education budgets are council tax
income and non-domestic rates income.3 A 2006 report of the National Assembly for
Wales School Funding Committee4 recommended that funding through the local government
revenue settlement be reviewed. This led to the commissioning of a report that investigated
alternative methodologies that could be used to predict relative demand/need for local
government services, particularly education, with an emphasis on the importance of outcomes-
based distribution formulas.5 Therefore, the present report aims to determine the link between
resource distribution and outcomes (such as cognitive attainment6) to best inform the new
student finance delivery formula.
A 2000 review of the UK literature by Vignoles et al. indicated that some educational resources
seem to have an impact on outcomes.7 That report cited a positive relationship between the
school-level pupil-teacher ratio and attainment, but noted that there is almost no UK evidence
that smaller class sizes lead to better outcomes. This review now requires updating, as recent
2 Outcomes in this review include pupil cognitive attainment, educational progression, and future wages. We also
include some outcomes that can be seen as both mediating variables and outcomes in their own right, such as school climate and educational aspirations. However, we do not include ―throughputs‖—variables that have little intrinsic value on their own, but rather have value from the way in which they mediate or moderate outcomes. A key example of such a throughput is teacher quality. 3 Welsh Assembly Government website. School funding. Accessible at
http://playlearngrowwales.gov.uk/topics/educationandskills/learningproviders/schools/schoolfunding 4 The School Funding Committee report can be found on the National Assembly for Wales website, see:
http://www.assemblywales.org/N0000000000000000000000000045329.pdf 5 Bramley, G., and Watkins, D. (2007). Alternative resource allocation models for local services in Wales. Final report for
the Welsh Assembly Government. 6 Attainment in this study is used to refer to cognitive attainment unless otherwise specified, and is typically
operationalised in the included studies as school grades or standardised academic test results. This is distinct from other forms of attainment measured in pupils such as affective attainment. It is also distinct from academic achievement, which often refers to educational progression—typically operationalised by measures such as high school completion or higher education enrolment. However, there is some inconsistency in the literature regarding the use of the term achievement. Where academic achievement is used in the included studies to denote school grades or test results as opposed to educational progression, it will be referred to as attainment in the present review to avoid confusion. 7 Vignoles, A., Levačić, R., Walker, J., Machin, S. and Reynolds, D. (2000). The relationship between resource
allocation and pupil attainment: A review (DfEE Research Report 228). London: DfEE.
REA: Educational resources and outcomes
Matrix Evidence | 20 January 2011 10
moves towards cost-effective education has prompted a surge in studies looking at the
association between school funding and educational outcomes.
Figure 1 shows the process model within which this report is couched. The model posits that
policy and funding decisions determine the allocation of funds. This in turn leads to an
investment in resources (e.g., reduced class sizes, better facilities, better trained teachers) or
investment in educational interventions (e.g., ICT training, reading interventions). It is then
hypothesised that these resources or interventions lead to an improvement in outcomes, such
as improved student attainment or decreased poverty. Some research projects test only the link
between policy decisions or allocation of resources and outcomes; these links are represented
by dotted lines in Figure 1 as they are considered to be indirect effects8 on the outcomes.
The present report focuses on the white boxes in Figure 19, although the various other paths will
be occasionally referred to throughout the report. As such, the key aims of this research are to:
Establish the evidence base for the effect of increased resource allocation on improved
outcomes (e.g., attainment).
Relate the findings from other educational systems to the Welsh context, and evaluate
their relevance.
Establish the most important outcomes through which an outcomes-based resource
distribution formula could be developed.
Figure 1. Process model of funding and outcomes within which the present report is
situated.
Note. Dotted lines represent hypothetical relationships that were not examined in this review. Grey boxes
indicate factors that were not included in the evidence review.
8 Indirect effects are those effects on a variable by another that are mediated by other (third) variables. For example,
school funding is likely to have an effect on achievement indirectly through the equipment or staffing improvements that the additional money can buy, rather than as a direct effect of the money itself. 9 Although they have been discussed in similar reviews on the topic (notably Vignoles et al., 2000), educational
interventions such as Reading Recovery programmes, will not be reviewed in this report. In the Welsh system, decisions to conduct such interventions lie within the school, in contrast to inputs such as per pupil expenditure that are largely under the remit of the Welsh Assembly Government. We have limited the review to those inputs that are more clearly under the influence of the WAG to enhance conceptual clarity and due to limitations of time; however, we do acknowledge the importance of educational interventions in broader debates about improving student outcomes.
REA: Educational resources and outcomes
Matrix Evidence | 20 January 2011 11
3.2 Research Context
The unit of analysis used in quantitative studies in this field of research is an important
consideration. Educational systems can be thought of as multilevel models, in which students
are nested within classes, classes within schools, schools within local education authorities
(LEAs), and LEAs within counties or even countries (Hox, 2003). Each level in the system has
different features that can be measured by certain variables. Some variables pertain to the
student level, such as individual attainment or aspirations to attend university. Other variables
are measured at the school level, such as the number of teachers in the school, or the type of
school (e.g., primary school, secondary school). Many studies aggregate (average) data
collected at lower levels in the hierarchy to form higher-level variables. The most common is
attainment: the educational attainment of individual students is commonly aggregated to obtain
estimates of school-level attainment. The problem with this is that aggregating data loses
important information about the variance between students within in a school. As a result, when
school level variables (e.g., the size of the school) are correlated with school-average
attainment, estimates of the relation may be biased. Thus, using aggregated data can reduce
our confidence in the findings of some studies—often referred to as aggregation bias.
The most appropriate method for dealing with aggregation bias is to use multilevel modelling,
which does not use aggregated data. Rather, multilevel modelling estimates the amount of
variance both within and between schools (i.e., taking into account variation between students
and schools), which allows for more accurate estimation of relationships (Goldstein, 2005). The
educational resourcing literature is rife with analyses based on aggregated data that do not use
this multilevel modelling approach (see Vignoles et al., 2000, for a review), and therefore need
to be interpreted with some caution. However, this is typically due to the way in which large-
scale datasets are created. Often only school- or district-level datasets are available, making
this problem difficult to overcome.
A further problem in this literature is the issue of endogeneity. In simple terms, this refers to the
problem stemming from the way in which resources are distributed to schools; i.e., resources
are likely to be distributed non-randomly across the school system. For example, Jenkins et al.
(2006) noted that, ―In the English school system this [endogeneity] could arise through the
compensatory funding mechanism whereby the per pupil funding received by a school is related
to indices of social deprivation such as free school meals (FSM) status, which are in turn
inversely related to pupil attainment‖ (p. 4). In other words, factors beyond the inputs and
outcomes under examination might be fundamentally interlinked with this relationship.
Endogeneity can be partly resolved by including additional explanatory variables in the model of
the effect of inputs on outcomes. For example, including student socioeconomic status (SES) in
the model can help to explain the non-random distribution of resources associated with
additional funding for low performing schools. Thus, studies that include various background
REA: Educational resources and outcomes
Matrix Evidence | 20 January 2011 12
variables (such as SES, community poverty, etc.) will generally allow greater confidence in the
findings relating inputs to outcomes.
Linked to this is the notion of value-added models, which look at incremental improvements in
attainment over time. Failing to statistically control for students prior attainment can lead to
inflated estimates of the influence of resourcing on subsequent attainment. Although many of
the studies in this review do include a measure of prior attainment in their models (particularly
the more recent ones), not all studies in this field take this into account.
Finally, it is important to note that many studies in this field of research implicitly adopt a
unidirectional approach, with the assumption that resourcing inputs lead to changes in pupil
outcomes. However, as Figure 1 suggests, it is possible that outcomes might drive the resource
allocation in such a way that outcomes lead to changes in the inputs. For example, low
achieving schools are often given financial boosters based on their low performance on
standardised tests. It is likely that a reciprocal effects model, in which resources affect
subsequent outcomes, which in turn affect future resource allocations, is likely to exist (see the
causal loop in Figure 1). Levačić (2007) refers to this as two-way causality10
.
3.3 Approach
Matrix Evidence has used rapid evidence assessment (REA) methods to search for and analyse
data on educational resourcing. The REA process consisted of four stages:
Searching using electronic databases and other sources of literature;
Screening the literature identified in the search stage for relevance, using inclusion
criteria that are clearly defined and set in advance of screening;
Assessing the literature for methodological quality and extracting data using
standardised forms; and
Synthesizing the data to provide an overview of research findings.
The REA method meets the research needs as it is less biased, more thorough and more
transparent than a traditional literature review. The emphasis on transparency and replicability
in the reporting process minimises threats of bias often associated with traditional reviews, and
therefore builds greater confidence in the research findings.
A guide to conducting REAs can be found on the UK Government Civil Service website (Civil
Service, 2009). Below is an overview of the methods employed by Matrix Evidence (see Figure
2). For full technical details related to this report, please see the Appendices.
10
Levačić, R. (2007). The relationship between pupil attainment and school resources. In T. Townsend (ed.), The
International Handbook on School Effectiveness and Improvement (Chapter 22, 395-410). Dordrecht: Springer.
REA: Educational resources and outcomes
Matrix Evidence | 20 January 2011 13
Figure 2. The REA workflow
3.3.1 Research questions
From the broader aims described above, the main research questions explored in this report
are11
:
1. What evidence is there that resources have a positive effect on outcomes such as
student attainment levels?
2. From this, what outcomes might be useful for consideration for inclusion in resource
funding formulae?
3. How relevant is the international evidence to the Welsh context?
3.3.2 Searching
We searched 10 electronic databases using three main clusters of search terms. A list of the
databases and other sources searched can be found in Section A.1 of Appendix A. The clusters
contained terms relating to
Schools and education
11
Taken from the tender specification document.
REA: Educational resources and outcomes
Matrix Evidence | 20 January 2011 14
Funding or resources
Outcomes
Further restrictions were placed on the searches where possible. These
were:
Date restriction ―1999 – Current‖ to maximise the relevancy of the
evidence
Restriction to English language and Welsh language abstracts
The initial draft of the search strategy was constructed at King‘s College,
London in consultation with Matrix Evidence and Professor Rosalind
Levačić of the Institute of Education. A brief scoping review of the topic area was conducted and
the strategy was further developed into a preparatory working draft. Relevant officials from the
Welsh Assembly Government were consulted on this draft and their suggestions were
incorporated into the strategy.
This draft was tested across a couple of days, first to ensure that the relevant material was
being tracked and secondly, to assess the volume of returns that the project could anticipate.
This volume was initially large, which is understandable given that broad, popular terms such as
―funding‖ and ―attainment‖ were included in the strategy. We reviewed the strategy further,
looking at the thesauruses in the Education Resources Information Centre (ERIC) and the
British Education Index (BEI) again to track our terms. In addition, we looked to localise terms
within clusters as well as rooting phrases using speech marks.
The final strategy was again reviewed by Levačić and the client before searching commenced.
The searching started towards the end of September 2009 using a selection of resources
selected by the Centre for Evidence and Policy in consultation with Matrix and our experts. It
was necessary to run a shorter version of the strategy in some databases that not permit longer
search syntax. The searching was completed by the middle of October 2009.
3.3.3 Screening
Through the searching process, we identified 3,752 hits. Screening was
conducted on the titles and abstracts of these hits by assessing each
record against the inclusion criteria, below.
Inclusion criteria
1. The study is available in English or Welsh language.
2. The study refers to the use or allocation of resources or funding.
3. The study refers to educational settings—specifically, primary
and/or secondary schools (i.e., educational institutions covering
students that were within the age range 4-16 years at the time of attending). Note that
this does not mean that the participants are limited to those age ranges, but rather that
the resource allocation referred to in (2) had to occur while the students were in schools
Systematic searching maximises the
likelihood of locating relevant evidence. It also helps to ensure that the sample of
studies is representative of the wider population of
studies that could be included in a
particular review.
Applying consistent inclusion criteria to
each piece of evidence reduces the potential for bias that
is associated with ‘hand-picking’ studies to include in a review.
REA: Educational resources and outcomes
Matrix Evidence | 20 January 2011 15
for 4-16 year olds. Thus, any resource allocation to universities, for example, would fail
this criterion.
4. The study includes the measurement of outcomes such as attainment, drop-out, school
exclusion, post-school participation in education/training.
5. The study provides quantitative or qualitative data on the outcomes mentioned in (4), or
is a systematic review (or meta-analysis) of the relevant literature. Non-systematic
literature reviews, ―think pieces‖, or policy pieces should not be included.
6. The study was published in 1999 or after.
7. The study was carried out in a context that ensures some level of transferability. Those
studies that, for example, are concerned with expanding the coverage of compulsory
education should be excluded. Studies from developing countries should also be
excluded.
3.3.4 Data extraction
The studies were data-extracted by two researchers. Piloting of the tool
ensured a good level of inter-rater agreement. We used a standardised data
extraction tool for each study to facilitate comparisons across different
studies (see tool in Appendix D). The tool included
Contextual data relating to the services provided, the sample and
population, and other relevant background information;
Data relating to the outcomes;
Methodological data on the aims and design of the study; and
Information on the methodological quality of the primary research and report.
3.3.5 Data synthesis
We brought together findings from the studies in a synthesis report. This involved grouping the
findings by outcomes. Within outcomes, the studies are presented by input type. UK studies
received higher precedence in the synthesis than international literature. The order of
presentation is summarised in Figure 3, below.
There are some limitations to the literature presented in this report that should be taken into
account. A methodological note regarding problems more generally with quantitative evidence
in the field of educational resourcing research can be found in Section C.2 of Appendix C. A
brief annotated bibliography of each of the included studies can be found in Section C.3 of
Appendix C. These should be considered when assessing the strength of the conclusions of this
report.
Using a data extraction tool
ensures that similar information is
extracted from each study. This facilitates comparisons across
studies.
REA: Educational resources and outcomes
Matrix Evidence | 20 January 2011 16
Figure 3. The structure of the research synthesis in the present report
Note. The studies in the report are not given a quantitative ―confidence‖ ranking. Rather, the
methodologically stronger studies (―high‖ confidence) are presented towards the start of each section,
while weaker studies (―low‖ confidence) are presented at the end of the section.
3.3.6 Summary
In summary, the method used involved searching, screening, data extraction, and synthesising
the research evidence. The methods are detailed in Appendix A, which enhances transparency
about the decision to include some evidence but not others. The emphasis in the synthesis on
confidence in the research findings and transferability to the Welsh context help to ensure that
the evidence reported herein is useful for its intended purpose.
3.4 Results
3.4.1 Studies located
The searches identified 3,736 documents, to which 16 references provided by the experts were
added. A diagram setting out the flow of literature through the review can be found in Appendix
B. After applying the exclusion criteria, we were left with 60 primary studies. Details on the
inclusion and exclusion figures broken down by each database can be found in Appendix A.
REA: Educational resources and outcomes
Matrix Evidence | 20 January 2011 17
3.4.2 Summary of key study characteristics
This section summarises the characteristics of the 60 primary studies on educational resourcing
that were included in the REA. A summary of each study is presented in Appendix E, and a note
on their methods are available in Section C.3 of Appendix C.
Searches were limited to literature published since 1999. Within this time frame, references
tended to be fairly recent, with more than half concentrated on the last five years (n = 32).
Regarding the country of study, most references came from the US (n = 30), followed by the UK
(n = 25). Four studies contained Welsh-specific data. Nine studies were carried out in urban
settings, and one in a rural setting. The rest of the studies were either a combination of rural
and urban sites or did not specify the type of location.
Educational inputs were mainly operationalised in the form of per-pupil expenditure, pupil-
teacher ratio, and grant funding12
. Educational outcomes included attainment (e.g., test scores),
educational progression (e.g., access to higher education), and truancy/attendance. Non-
educational outcomes tackled in the included studies covered future wages, aspirations and
crime-related behaviour.
3.4.3 Resources and student attainment
Fifty-three of the 60 studies included in this review included some measure of student
attainment. This could either be in the form of school grades or performance on a standardised
test. Below, we provide a brief summary of each study included in the review, organised by the
resource type. Within each subsection of the review of attainment evidence, the studies are
roughly presented in descending order of methodological quality (i.e., the strongest studies are
presented first). Table 1 provides a summary of the age range (e.g., primary school students)
and level of data (e.g., pupil-level data) for the 39 studies the included specific resource
allocations in this review; the remaining 14 studies are discussed later.
12
Grant funding refers to general grants and funding initiatives that are not measured by a specific variable (such as
PTR); rather they measured student attainment before and after a new funding policy or initiative was introduced.
REA: Educational resources and outcomes
Matrix Evidence | 20 January 2011 18
Table 1
Country, school age of pupils, and data level for the 39 studies examining specific resourcing
variables with attainment
Citation Country School age of pupils Data level
Addonizio (2009) USA primary school level
Babcock & Betts (2009) USA primary school level
Betts & Morell (1999) USA secondary school level
Betts et al. (2003) USA primary, middle, and secondary
pupil level
Blatchford et al (2002) UK - England primary pupil level
Blatchford et al (2004) UK - England primary pupil level
Blatchford et al (2007) UK - England primary pupil level
Darling-Hammond (2004)
USA secondary district level
Dustmann et al. (2003) UK - England and Wales
secondary pupil level
Fuchs & Wößmann (2007)
31 countries secondary pupil level
Graddy & Stevens (2005)
UK private secondary school level
Greene et al. (2007) USA secondary school level
Grubb (2008) USA secondary school level
Haimson (2000) USA primary school level
Holmlund et al. (2008) UK - England primary pupil level
Hoxby (2004) USA primary and secondary district level
Iacovou (2002) UK primary pupil level
Ilon & Normore (2006) USA primary school level
Jacob (2003) USA primary school level
Jenkins et al. (2006) UK - England secondary schools school level
Koshal et al. (2004) USA high school district level
Leuven et al. (2007) the Netherlands primary pupil level
Levacic & Hardman (1999)
UK - England and Wales
primary and secondary school level
Levacic & Marsh (2007) UK - England secondary pupil level
REA: Educational resources and outcomes
Matrix Evidence | 20 January 2011 19
Citation Country School age of pupils Data level
Levacic et al. (2005) UK - England secondary pupil level
Loubert (2008) USA elementary school level
Machin et al. (2006) UK - England primary and secondary district level
Michie (2009) USA elementary and secondary
school level
Miller-Whitehead (2003) USA elementary unclear
Muijs & Reynolds (2003)
UK - England primary school level
Payne & Biddle (1999) USA secondary district level
Pugh et al. (2008) UK - England secondary school level
Ram (2004) USA elementary and secondary
state level
Richmond (2008) USA middle school level
Steele et al. (2007) UK secondary pupil level
Waldfogel & Zhai (2008) 7 countries preschool and primary pupil level
West et al. (2001) UK - England secondary district level
Wilson et al. (2002) UK - Scotland primary school level
Wößmann (2001) 39 countries secondary pupil level
REA: Educational resources and outcomes
Matrix Evidence | 20 January 2011 20
Expenditure per pupil
Of the 54 studies reporting student attainment as an outcome, 23 examined
per-pupil expenditure as a predictor. Of these, eight reported on UK data.
Overall, the eight UK studies indicated that expenditure had a small but
positive association with attainment. The remaining 15 studies were based
on international data (mostly from the US). The international findings were
more mixed than the UK literature, with some studies even reporting small
(albeit generally non-significant) negative relations with attainment. In
general, across all 23 studies, the effects tended to be more positive for
maths and science attainment than for English/verbal attainment or general
school grades. Details of the studies are below
UK studies
Eight studies reported data from UK students. In general, these showed a
positive relation between per-pupil expenditure and attainment.
In one of the better quality studies, Levačić et al. (2005) examined the
relation between educational inputs and students‘ attainment at Key Stage 3
in English secondary schools. They found that higher levels of expenditure
per pupil were associated with higher test scores in maths and science, but
not in English; these effects were small. This finding was echoed by Steele
et al. (2007); they reported a small but statistically significant increase in
mathematics and science attainment with increased expenditure per pupil in
UK secondary schools, while expenditure had a slightly negative effect (yet
significant) relation with English attainment.
In combination, these two high-quality studies suggest that maths and science attainment could
be significantly affected by increased expenditure, whereas English might not be linked to
funding. Looking at final school grades more generally (rather than subject-specific attainment),
Jenkins et al. (2006) investigated the impact of additional school resources on pupil attainment
in the GCSE scores. They used data from the 2003 National Pupil Database, and found that
small gains in attainment were associated with higher per-pupil spending.
Levačić and colleagues have examined the expenditure issue in relation to specific school
types. Levačić and Marsh (2007) focused on secondary modern schools in England. They used
a national data set at the pupil level for GCSEs sat in 2001. They found that modern secondary
schools, which typically had lower per-pupil expenditure than comprehensive or grammar
schools, had lower average attainment. This follows the general trend of the Jenkins et al.
(2006) study, but it is difficult to isolate the effects of the school type from the effects of
additional funding. In contrast, Levačić and Hardman (1999) compared the performance from
1991 to 1996 of grant-maintained (GM) and local educational authority (LEA) primary and
secondary schools in England and Wales. They reported that the average real school budget
per pupil had a significant negative effect on GCSE scores, for both GM and LEA schools. The
authors suggested that this counterintuitive finding might be the result of greater funding being
In general, the relation between per
pupil spending is more positive for
maths and science achievement than
English/verbal achievement and general school
grades.
UK studies tend to support a small,
positive association between per-pupil expenditure and
attainment—depending on the subject. However, one study found a small, significantly
negative relation with overall grades.
REA: Educational resources and outcomes
Matrix Evidence | 20 January 2011 21
allocated to schools with a greater proportion of special educational needs students or more
free school meals, whose students tend to also be academically disadvantaged. However, this
study did not have pupil level data and did not sufficiently control for pupil background effects.
Holmlund et al. (2008) reported a positive and significant effect of school expenditure on
English, maths, and science attainment at Key Stage 2. However, school and pupil deprivation
influenced this relation. Increases in funding have a greater effect on English in deprived
schools, while the opposite is true for science attainment.
Graddy and Stevens (2005) differed from most of the other studies in the review by focusing on
UK independent schools. They found that capital spending per pupil had a statistically
significant positive effect on the proportion of A-grades that a school received. In another study
focusing on the proportion of certain GCSE grades attained, this time in state schools, West et
al. (2001) analysed data on 96 LEAs in England. Spending per pupil was positively associated
with national test and examination performance after statistically controlling for poverty. This
relationship was not statistically significant for Key Stage 1 results but was a significant
predictor of the number of students with five or more A*-C grades at Key Stage 4 (GCSE) level.
Pugh et al. (2008) analysed school-level performance at GCSE in England. They found a small
but significant positive effect of per-pupil spending on both the average points score and the
number of GSCE grades at A*-C level. They observed that the estimated effect of expenditure
in specialist schools was lower than in non-specialist schools.
In summary, UK evidence suggests that maths and science attainment, and the proportion of
students receiving A*-C levels at GCSE, are positively related with per-pupil expenditure.
English attainment does not appear to be related to expenditure.
International studies
Fifteen studies were located that used non-UK data. Overall, these also
supported the positive link between expenditure and attainment.
The strongest international studies come from analyses of large international
databases. Fuchs and Wößmann (2007) conducted an analysis of data from
31 countries using the OECD‘s Programme for International Student
Assessment (PISA) 2000 dataset of 15-year-old students. They found that
educational expenditure per student was strongly and positively correlated
with maths test results, but was not significant for science and reading
attainment. In contrast, Wößmann‘s (2001) study reported analyses on data
from 39 countries, with the aim of estimating the effects of family background, resources and
institutions on mathematics and science secondary school performance. They concluded that
the estimated relationship between the overall amount of resources and student performance
was not significantly positive. Interestingly, it was observed that school autonomy regarding
purchasing supplies positively predicted student attainment.
Focusing on a younger age group, Waldfogel and Zhai (2008) looked at maths and science
outcomes for students in seven countries. They found that public preschool expenditures have
The international literature also
indicates that the positive association between per-pupil expenditure and attainment might depend on the
subject.
REA: Educational resources and outcomes
Matrix Evidence | 20 January 2011 22
small but statistically significant positive effects on fourth graders‘ science scores, while
expenditure in primary education has a smaller but significantly negative effect on subsequent
attainment. These findings indicate that the level of schooling might be important in determining
the impact of expenditure on attainment. Importantly, they also noted that children from low-
resource homes might gain more from increased public preschool expenditures than other
children.
Hoxby (2004) analysed 30 years of national school-level data from the US. They found that test
scores have remained relatively unchanged over that period, but expenditure per-pupil had
doubled. This led Hoxby to argue that educational productivity had decreased, which explains
the absence of improvement despite the injection of funds.
Darling-Hammond (2004) explored disparities in educational access in the US states of South
Carolina13
and Massachusetts. They analysed existing school district data and found that
Massachusetts‘ net school funding was not a significant predictor of English results once race,
socioeconomic status, and teacher characteristics were taken into account. In contrast, school
spending was a significant predictor for mathematics attainment. Similarly, Ram (2004)
conducted an analysis of elementary and secondary schools using national data in the US.
Expenditure had a significantly positive relation with maths scores and SAT scores, but a
negligible (non-significant) positive effect on verbal scores. These findings reflect that of the UK
literature; that is, the school subject is an important factor. However, in contrast, Richmond
(2008) did find a positive association between per-pupil expenditure and verbal student
attainment. A significant relation was also found with maths attainment for middle school
students in the US state of Virginia.
Several studies have found a small but positive relation for secondary students in the US.
Koshal et al. (2004) examined data from 576 school districts in Ohio. They found that
expenditure per pupil was linked to small but ―practically meaningful‖ increases in attainment.
Jacob (2003) found that both total expenditures and the fraction spent on instruction are
positively related to student mathematics attainment in Chicago, although the magnitude of the
effects was quite small. Payne and Biddle‘s (1999) study aimed to evaluate the relation between
poor school funding, child poverty, and maths attainment for eighth grade students in 32 states
of the US. They found that district-level per-student funding was a significant positive predictor
of maths attainment, even after controlling for child poverty, ethnic minorities, and average level
of curricular instruction.
In opposition, further studies on primary school students in the US are not as promising.
Addonizio (2009) found a significant positive relation between expenditure and attainment for
Grade 5 students but a non-significant relation for Grade 3 students in Minnesota. Similar to the
Waldfogel and Zhai (2008) study above, Ilon and Normore (2006) found that expenditures per
student are negatively related to attainment in elementary school students (this study was
conducted in Florida).
13
This variable was not included in the South Carolina models because adequate data were not available.
REA: Educational resources and outcomes
Matrix Evidence | 20 January 2011 23
Other studies on elementary school students have found small or non-significant findings.
Loubert (2008) analysed the data from a natural quasi-experiment. They examined elementary
school attainment before and after increases in funding occurred in Dallas County, Texas in
1993. They reported that overall school attainment increased slightly with increased funding,
although it was not clear if the results were statistically significant or not. Miller-Whitehead
(2003) reported a study on Grade 5 students in Tennessee. They noted that per-pupil spending
was not a significant predictor of science attainment.
In summary, the international literature generally supports a small positive association between
per-pupil expenditure and student attainment in secondary schools. The relation is weaker—or
even negative—in younger grades. This is perhaps because of the different resource needs of
primary school students14
. Also, a difference in subject-specific attainment gains was observed,
in that some studies found no support for improvement in English but did for maths attainment.
Class size
The relation between class size and attainment has received mixed support.
Nine studies reported on the relation between class size and attainment.
Three of these were reviewed for the per-pupil expenditure, above. Fuchs
and Wößmann (2007), a relatively high-quality study, reported a positive
relation between class size and attainment in maths and science, but no
relation with reading attainment. That is, as class size increased, so did
attainment in maths and science. Miller-Whitehead (2003), a poorly reported study, found that
class size significantly and positively predicted Grade 5 science attainment. Ilon and Normore
(2006), a weaker study, reported a negative relation between class size and attainment.
In one of four UK studies on the relation between class size and attainment, Dustmann et al.
(2003) found a significantly negative relation between class size and exam results (O-levels) in
English and Welsh secondary school data. Iacovou (2002) found that smaller class size has
positive effect on reading at ages 7 and 11 when school size and type were controlled for, but
found no effect on maths attainment. However, ―small‖ class sizes at the period in which this
study was conducted would be large by today‘s standards: approximately 80% of the students
were in classes of more than 30 students.
Blatchford et al. (2002) found positive effects of smaller classes for literacy and numeracy
during the first year of schooling. In an important demonstration of the need to include prior
attainment in the modelling, they reported the strongest effects were found for pupils who had
the lowest baseline scores on school entry. Blatchford et al. (2004), found no effects of smaller
class size on year 4 or 5 maths and literacy attainment, or year 6 science and maths attainment.
However, contrary to expectations, a positive relationship between class size and literacy was
found in year 6.
14
This should be considered in comparison with the UK literature, which found mixed and occasionally positive relation
between resourcing (class size) and attainment.
The findings for class size and attainment
are mixed and inconclusive.
REA: Educational resources and outcomes
Matrix Evidence | 20 January 2011 24
Similar to Blatchford et al. (2004), Betts et al. (2003) reported no significant relation between
class size and attainment in high schools in San Diego in the US. In the same study, they also
looked at elementary and middle schools, with different findings. In elementary schools, when
controlling for teacher characteristics, they found a significantly negative relation between class
size and reading attainment, but no relation between class size and maths attainment. For
middle schools, they found a significantly positive effect of class size on maths attainment, but
no significant relation with English attainment. Babcock and Betts (2009) also examined
elementary schools in San Diego. However, they found that there were generally no significant
relations between class size and maths or reading attainment.
The findings for class size are inconsistent, and so firm conclusions cannot be drawn. However,
there appears to be a trend across studies towards higher attainment in smaller classrooms.
Pupil-teacher ratio
Fourteen studies report the relation between pupil-teacher ratio (PTR) and
attainment. Ideally, the relation should be negative. This means that as the
ratio of students to teachers decreases (that is, there are fewer students per
teacher) then attainment increases.
UK studies
Four of the 14 studies report UK data. Overall, the UK data indicate that
lower PTRs might be linked with higher attainment, depending on the
subject.
Graddy and Stevens (2005) found a small negative relation of PTR with
GCSE scores; Steele et al. (2007) reported a negative relation with maths
and science but not English attainment; Jenkins et al. (2006) found a
negative relation between PTR and science but not with maths or English; and Levačić et al.‘s
(2005) study mirrored those of Jenkins et al. (2006). In summary, the UK studies seem to
indicate some negative relation between PTR and attainment for science and perhaps maths,
but not English.
International studies
Ten international studies examining the relation between PTR and attainment are included in
the review. These almost unanimously conclude that there is no significant relation between
PTR and the attainment measures considered.
Five of the ten international studies have already been discussed above (Addonizio, 2009;
Darling-Hammond, 2004; Jacob, 2003; Richmond, 2008; Wößmann, 2001). These five studies
found no or mixed support for the relation between PTR and attainment. The further five
international studies are described below.
Two further international studies found no relation between PTR and attainment. Betts and
Morell (1999) examined the secondary school characteristics that are determinants of college
In the UK, decreases in the pupil-teacher
ratio (PTR) appear to be related to
increases in science and possibly maths attainment, but not
English attainment. In the international
literature, research tends to suggest no
relation between PTR and attainment.
REA: Educational resources and outcomes
Matrix Evidence | 20 January 2011 25
grade point average in San Diego. They reported no significant relation between PTR and
subsequent attainment at college. Greene et al. (2007) analysed data from secondary schools
in New Jersey in the US. Class size, student-teacher ratio and student-aide ratio were not
significant predictors of grades in languages, arts or mathematics.
Leuven et al. (2007) compared the effects of two subsidies (more personnel in the school and
more funding for ICT equipment) targeted at schools with large proportions of disadvantaged
pupils on attainment in the Netherlands. They concluded that both resources led to ―harmful‖
outcomes.
Two studies found some benefits of reduced PTR on attainment, although both of these studies
were quite weak. Grubb (2008) examined the effectiveness of an ―improved‖ school finance
system on secondary school attainment using a United States nationwide dataset. More pupils
per teacher were associated with reduced maths scores; other attainment measures were not
associated with PTR. Haimson (2000) assessed a class-size reduction programme on teacher
perceptions and student performance after one year in New York. Unlike most of the other
studies, Haimson focused on interview data and reported that principals and teachers felt
attainment had improved. However, as this was based on teacher perception as opposed to
actual attainment on tests or school grades, this finding should be viewed very cautiously.
Facilities/books
Three studies examined the effects of improved facilities or library resources
on attainment15
. The results are mixed, but indicate that there might be some
improvement in attainment with improved ICT or library facilities.
Machin et al. (2006) examined English data to test the impact of ICT
investments in student attainment at primary and secondary levels. They
found that ICT funding led to significant improvements in English and science
test scores—but not maths—at age 11. In contrast, the Dutch study by
Leuven et al. (2007) found that funding for improved ICT facilities had no
effect on attainment.
Michie (2009) analysed data from 24 US states to evaluate the Improving Literacy Through
School Libraries programme in 2005. That study suggested that an increase in the number of
books in a school might increase attainment.
15
Other studies were located that focused on ITC or reading training programmes, which are not reviewed here. The
included studies differ in that the main focus was on how increased funding on these facilities improved attainment, rather than on the effectiveness of an intervention.
The evidence suggests that there might be a positive
effect of improved ICT and library facilities
on achievement, although findings are
mixed.
REA: Educational resources and outcomes
Matrix Evidence | 20 January 2011 26
General grants and funding initiatives
Fifteen studies included in the review were based on grants and funding
initiatives. These differed from the abovementioned studies in that they were
not measured by a specific variable (such as PTR); rather they measured
student attainment before and after a new funding policy or initiative was
introduced. Because of the heterogeneity in the range of initiatives
introduced, we present the aims of the 14 studies along with the features of
the grant and the attainment outcomes in Table 2. Overall, the variety of initiatives seemed to
be associated with small but positive changes in student attainment.
Grant initiatives appear to be
associated with small but positive effects on
attainment.
REA: Educational resources and outcomes
Matrix Evidence | 20 January 2011 27
Table 2
Description of 14 studies reporting the effects of grant/initiatives on attainment
Citation Country School level Study aim Features of the grant/ initiative
Attainment outcomes
Bradley & Taylor (2002) UK - England
secondary To investigate the effects of the quasi-market on efficiency and equity in the secondary education sector in England during the 1990s
Funding, directly related to age-weighted pupil numbers and control over financial resources which has been devolved from the local education authority to the schools themselves
Positive effect on exam results
McNally (2005) UK - England
secondary To evaluate the Pupil Learning Credits (PLC) policy initiative
Schools with more than 50% of pupils known to be eligible for Free School Meals obtain extra funding
PLC increases maths performance, but has no effect on English attainment
Noden & Schagen (2006)
UK - England
secondary To evaluate a Specialist Schools programme
Additional (one-off) capital funding and recurrent grants for four years
GCSE results is positively and significantly correlated with the Specialist Schools programme.
PricewaterhouseCoopers (2003)
UK - England
primary and secondary
To investigate the effects of both family and school capita on math and reading test scores for a sample of elementary and middle-school age children
Funding In relation to primary schools, capital investment had some positive effect on attainment. For secondary schools, the capital investment was also positive.
REA: Educational resources and outcomes
Matrix Evidence | 20 January 2011 28
Citation Country School level Study aim Features of the grant/ initiative
Attainment outcomes
Tikly et al. (2002) UK - England
primary and secondary
To evaluate the effectiveness of the Ethnic Minority Achievement Grant (EMAG)
EMAG grants: mentoring, administrative support, and monitoring
The overall improvement rate from 1998-2000 in the proportion of pupils achieving Level 4 and above for the 81 LEAs which provided data for the three years was 11.9 percentage points in English and 14.9 percentage points in mathematics.
Estyn (2008) UK - Wales primary and secondary
To review the impact of funding with a particular focus on how effective it was in helping schools and local authorities to tackle the underachievement of pupils with socio-economic disadvantage
Funding, spent in several different ways
Positive effect on GCSE results
AEL Inc (2000) USA primary To evaluate the impact of 10 years of the Kentucky Educational Reform Act
Improved school facilities, new support programs, technology resources, and educational materials
Some positive improvement in test scores
Bacolod et al. (2007) USA primary and middle
To examine California's accountability system; focused on how schools used financial rewards
Governor‘s Performance Award: money awarded to schools (per pupil) for good performance on standardised tests
Little improvement in test scores or other measures of attainment
REA: Educational resources and outcomes
Matrix Evidence | 20 January 2011 29
Citation Country School level Study aim Features of the grant/ initiative
Attainment outcomes
Baskett et al. (2004) USA kindergarten To test whether full-day or half-day kindergartens are associated with improved educational and development scales
Half-day versus full-day kindergarten
Mixed results on reading, literacy, and maths attainment
Cramer (2006) USA secondary To assess the impact of smaller learning communities on student outcomes
Grants provided to implement smaller learning communities
Some positive effect on academic attainment
Kinnucan et al. (2006) USA ns To examine the relationship between state aid and student performance
Funding Increased state aid can have important benefits for students if directed toward improved teacher pay and not to reductions in class size.
Klaauw (2008) USA elementary and middle
evaluation of the impact of Title I funding of compensatory education programmes on school finance and student performance in New York City public schools
Funding Title I has not led to better student outcomes
Parcel & Dufur (2001) USA primary To evaluate a Specialist Schools programme
Additional (one-off) capital funding and recurrent grants for four years.
School financial capital has a positive effect, with children who attend schools with higher per-pupil expenditures experiencing increases
REA: Educational resources and outcomes
Matrix Evidence | 20 January 2011 30
Citation Country School level Study aim Features of the grant/ initiative
Attainment outcomes
in math attainment
Reynolds et al. (2001) USA primary To conduct cost-benefit analysis of a federally financed, comprehensive early childhood programme (The Title I Chicago Child-Parent Centers)
Educational and family support services to low-income children
Participation in the programme led to significantly better school attainment in most measures analysed.
REA: Educational resources and outcomes
Matrix Evidence | 20 January 2011 31
3.4.4 Resources and other outcomes
In total, 17 of the 60 studies examined the relation between resources and non-attainment
outcomes. These studies were diverse in both the educational inputs measured and the
associated outcomes. We grouped the outcomes into six categories:
Educational progression . This is typically measured either by participation in school
after compulsory education finishes, completion of non-compulsory education or
enrolment at a higher education institution.
Truancy/ attendance. Truancy is measured as the number of unauthorised absent days,
while attendance refers to the number or proportion of days attended in the school year.
Future wages. Typically measured as post-school or post-higher education earnings.
Aspirations. In this review, two studies examined educational aspirations (i.e., the desire
to pursue higher education) and one measured occupational aspirations (i.e., the desire
to pursue professional careers).
Crime/ behaviour. These outcomes were a mixed bag of variables such as incarceration
risk, juvenile delinquency and disciplinary referrals.
Social outcomes. Again, this was a mixture of measures including the student‘s social
function, teacher morale, and school climate.
Table 3 overleaf summarises the studies by outcome. Studies marked as ―good‖ indicate that
increased resources were linked to improvements in the specific outcome. Studies marked as
―neutral‖ found no significant relation between resourcing and outcome. Studies with ―mixed‖
results were internally inconsistent; e.g., positive impact for some students but not others. The
one study marked as ―bad‖ had a negative effect of increased resourcing on the outcome. The
studies are detailed below, with reference to the resourcing type (e.g., per-pupil expenditure).
Educational progression
The UK literature is supportive of the relation between resourcing and educational
progression—regardless of the type of resourcing—but the data from the US suggest less of a
link.
Dustmann et al.‘s (2003) analysis of English and Welsh data suggests a
small but negative effect of class size on the decision to continue with
post-16 education after controlling for other variables. In a study of a
Scottish high school, Brown (2004) found that grant money improved rates
of higher education access. That is, more students progressed to further
education or training after the school received the grant funding. McVicar‘s
(1999) study on high schools in Northern Ireland found that the reduced
PTR and higher expenditure both have significant desirable impacts on staying at school. From
the four studies reviewed here, it seems that UK studies support the impact of resourcing on
improved educational progression.
The UK evidence suggests that educational
progression might be improved by
increased resourcing.
REA: Educational resources and outcomes
Matrix Evidence | 20 January 2011 32
Studies from the US are less promising. Betts (2001) measured educational
progression as the latest observation available on years of schooling
obtained for each woman in the US sample. High-school resources as
indicated by pupil-teacher ratio, spending per pupil, teachers‘ salary and
books per student did not significantly predict educational progression.
Cramer (2006) found that differences in drop-out rates and preparation for
post-secondary education were not statistically significant after high schools
received federal grant money. Grubb (2008) found that high PTRs reduce
the likelihood of completing a standard academic programme and of continuing to a four-year
college; however, it also increases the likelihood of graduating and going to a community
college. More encouragingly, grant-funded participation in a Chicago programme was
significantly associated with higher rates of high-school completion in Reynolds et al.‘s (2001)
study. These four US studies indicate no or mixed evidence of a relation between educational
resourcing and educational progression.
There is little or no evidence in the US
literature that educational
progression might be improved by
increased resourcing
REA: Educational resources and outcomes
Matrix Evidence | 20 January 2011 33
Table 3
Relationship between educational resources and various outcomes by study
Outcomes
Study
Educational
progression
Truancy/
attendance
Future
wages Aspirations
Crime/
behaviour Social
AEL Inc (2000) Improved
Anderson et al. (2008) Mixed
Arum & LaFree (2008) Improved
Baskett et al. (2004) Improved
Betts & Roemer (2005) Neutral
Betts (2001) Neutral Mixed
Bradley & Taylor (2002) Declined
Brown (2004) Improved
Cramer (2006) Neutral
Dustmann et al. (2003) Improved Improved Improved
Greene et al. (2007) Neutral
Grubb (2008) Mixed Improved
Haimson (2000) Improved Improved
Jones et al. (2006) Improved
McNally (2005) Neutral
McVicar (1999) Improved
Reynolds et al. (2001) Improved Improved
Note. ―Improved‖ indicates increased resources were linked to improvements in the outcomes; ―neutral‖ indicates no
significant relation between resourcing and the outcome; ―mixed‖ indicates that some of the finding supported the
relationship; ―Declined‖ indicates that resources had a negative relation with the outcome.
Matrix Evidence | 20 January 2011 34
Truancy and attendance
Only three studies assessed the benefits of resourcing on reducing truancy or
improving attendance. They were generally in favour of increased resourcing.
Dustmann et al. (2003) found that an increase in class size increases the
probability of truancy in England and Wales. In an English sample, McNally
(2005) found that reduction in absences after the introduction of the Pupil
Learning Credits (PLC) policy initiative was quite small. This could be
because the programme simply involved providing money to the school,
without any obvious change in factors that might improve attendance (such as student
incentives).
In the US, Jones et al. (2006) found that Texan high schools had a small but possibly practically
important change in daily attendance with the change in class size. Larger enrolments led to
lower attendance rates; this is similar to the English and Welsh findings of Dustmann et al.
(2003).
Future wages
Overall, it is not clear that high-school resourcing improves the chances of
higher post-secondary wages. However, only three studies examined this
relation.
Dustmann‘s (2003) analysis of English and Welsh data was promising: The
indirect effect of class size on wages (mediated by ―staying on decisions‖)
was negative. In other words, larger classes lead to a slight decrease in future
wages. In a national survey in the US, Betts and Roemer (2005) found only negligible increases
in subsequent wages with increased high school spending. In a mixed result of US data, Betts
(2001) found that white women's future wages were positively related to the high-school PTR
and negatively to books per student, while black women's wages were negatively related to the
PTR and positively related to books per student.
Aspirations
There are only two studies that refer to the effects of resourcing on
educational and occupational aspirations. Given their different findings, it is
not possible to glean whether aspirations are enhanced by increased
resourcing. Greene et al.‘s (2007) study of New Jersey high school students
did not find a significant effect of resourcing on educational aspirations.
Specifically, class size, student-teacher ratio, and student-aide ratio were not
significant predictors of aspirations to attend college. In contrast, Grubb‘s
analysis of national US data found that teacher salaries enhanced both
occupational aspirations and the intention to continue in schooling.
The small number of studies examining
truancy and attendance seem to indicate the benefits
of increased resourcing.
From the three studies reviewed
here, there is no clear relation between
resourcing and future wages.
No conclusions can be drawn regarding the relation between
resourcing and aspirations from the
two studies considered here.
Matrix Evidence | 20 January 2011 35
Crime and behavioural outcomes
All three studies measuring crime and behavioural outcomes focused on US
data. They each suggested benefits of resourcing for improving these
outcomes.
Haimson (2000) interviewed teachers and principals in New York elementary
schools that had received funding to reduce the PTR through reduced class
size and/or ―floating teachers‖. They reported that disciplinary referrals
dramatically reduced in the first year of the programme. However, this study shold be given low
weighting as it was based on teacher perceptions as opposed to actual attainment on tests or
school grades. Also examining the PTR, Arum and LaFree‘s (2008) US national survey found
that students from schools with lower PTRs have a lower risk of incarceration. However, they
also noted that states that invest more in reducing the PTR also tend to spend more on social
welfare programmes, which could explain some of this variation. In another project that could be
evidence of co-varying social phenomena, Reynolds et al. (2001) measured the impact of
educational and family support services to low-income children in Chicago. They reported
improvements in juvenile delinquency and child maltreatment rates.
Thus, improved resourcing seems to improve crime-related behavioural outcomes in US
studies. However, these findings are likely highly interrelated with other social services.
Social outcomes
Studies reporting social outcomes are mixed in terms of both the outcomes
measured and the impact of resourcing. It is difficult to draw firm conclusions
from these studies.
Bradley and Taylor (2002) reported an English study in which funding reform
was directly related to age-weighted pupil numbers. Control over financial
resources was also devolved from the local education authority to the
schools. They concluded that these reforms increased competition between
schools. Although this increased attainment (reviewed above), they also
cited evidence of a widening gap in the social composition of schools. Haimson‘s (2000)
interviews with teachers and principals of New York elementary schools indicated positive
perceptions of reduced class sizes. Specifically, they report that smaller classes led to higher
morale among teachers, less staff turnover, improved parent-teacher relationships, and more
collaboration between teachers.
Another positive finding came from a study in New Zealand. Anderson et al. (2008) reported
that additional funding for school nurse and social worker services improved school climate.
This is likely to be a mediating variable between resourcing and final outputs such as
attainment, although it can be argued that school climate is a desirable outcome in its own right.
The risk of committing future crime
behaviours might be reduced by increasing
school resources.
No conclusions can be drawn regarding the relation between resourcing and social outcomes, in part due
to the diversity of social outcomes
explored.
Matrix Evidence | 20 January 2011 36
On a staff school climate survey, significant positive increases were found on scales including:
Improvements in Last 12 Months, Support for Ethnic Diversity, Innovation Culture, and Students
are Friendly. In the student school climate survey, year 10 students showed more positive mean
scores than students in comparison schools on the three scales of Satisfaction with School,
Support for Achievement, and Support for Ethnic Diversity. Overall, these indicate some
improvements in school climate for both staff and students.
With an emphasis on parental perceptions, AEL (2000) reported on the effects of funding
increases in Kentucky. According to the authors, these funds improved school facilities, and
introduced new support programs. After the reforms were introduced, AEL reported that parents
felt the schools were preparing students for adult life. However, given that these are perceptions
of improvement rather than more objective measures, it is difficult to determine how
substantively meaningful such improvements are.
In a different type of resourcing issue, Baskett et al. (2004) compared half-day with full-day
kindergartens in Maine, USA. They reported that social functioning had small but significant
differences in favour of students in full-day kindergarten programmes.
In summary, the diversity of outcomes makes it difficult to review these studies. The findings are
quite mixed, although this could be due to the range of inputs and outcomes. Promisingly, some
studies did indicate positive social outcomes.
3.4.5 Summary of the findings
Summary of findings in relation to educational attainment
Fifty studies in the review examined attainment as an outcome. UK studies tend to support a
small, positive association between per-pupil expenditure and attainment. However, both the
UK and the international literature indicate that the positive association between per-pupil
expenditure and attainment might depend on the school subject. In general, the relation
between per-pupil spending is more positive for maths and science attainment than for
English/verbal attainment and general school grades.
The findings for the relation between class size and attainment are mixed. In the UK,
decreases in the pupil-teacher ratio (PTR) appear to be related to increases in science and
possibly maths attainment, but not to English attainment. However, in the international literature,
research tends to suggest no relation between PTR and attainment.
There might be a positive effect of improved ICT and library facilities on attainment, although
findings are mixed. Grant initiatives, in which funding is dedicated to broad improvement
programmes, appear to be associated with small but positive effects on attainment.
Overall, the studies indicated a small positive relation between educational resourcing and
attainment. However, the strength of the relationship depends largely on the school subject
Matrix Evidence | 20 January 2011 37
being measured: resourcing appears to have a stronger relation with science and maths
attainment than with English or verbal attainment. Still, the relation between attainment and
resourcing needs to be considered in the context of several student background and
instrumental variables. In particular, student prior attainment and socioeconomic status, and the
characteristics of the teachers (qualifications etc.) are important.
Summary of findings in relation to other outcomes
The evidence for non-attainment outcomes was sparser. This makes it harder to draw firm
conclusions. The four UK studies on educational progression (e.g., continuing to higher
education) suggest that it might be improved by increased resourcing. However, there is little or
no evidence for this from the four US studies.
The small number of studies examining truancy and attendance seem to indicate the benefits
of increased resourcing. Also with a small number of studies in the review, it appears that the
risk of committing crime-related behaviours (e.g., juvenile delinquency) might be reduced by
increasing resources—particularly reducing the pupil-teacher ratio.
The remaining outcomes examined here did not yield sufficient evidence. In particular, from the
three studies reviewed, there is no clear relation between resourcing and future wages. In
addition, no conclusions can be drawn regarding the relation between resourcing and
aspirations, or resourcing and social outcomes.
3.5 Implications
This REA was conducted according to systematic review principles. However, it was completed
in a short time and, because we limited the number of databases searched, we did not search
the literature in a comprehensive fashion. No authors were contacted for their data, and we did
not carry out any re-analysis of published data. However, the use of specific search strategies
and well-defined inclusion criteria ensured that no relevant study that we found was excluded
with the intention to support a specific hypothesis. Therefore, as much as possible, this work is
free from bias.
The ultimate aim of this exercise is to establish the link between resourcing and outcomes in
order to support the use of outcomes in determining a funding formula (see Figure 4). The
above research indicates that, to some extent and for some inputs and outcomes, increases in
resources can lead to improved outcomes. However, many of the studies above are
correlational in nature, such that they determine whether high levels of funding are associated
with higher levels of an outcome. This does not allow causal attributions to be made. In other
words, very few of the studies included here allow us to definitively say that increased resources
cause improved outcomes. As with any evidence synthesis, our findings reflect what is available
in the literature and the specific factors that primary study authors have chosen to investigate,
and so stronger causal relations can be established only by conducting more primary research.
Matrix Evidence | 20 January 2011 38
Figure 4. A hypothesised link of this research and the application to a funding formula
Note. S = the numerical strength of the relation between funding and outcomes, as per the Bramley and
Watkins (2007) report.
The studies included in the REA are heterogeneous in terms of their setting and context and the
particular factors investigated. It can be problematic taking the evidence from one context and
applying it to another context, as the same findings might not apply. In the present study, we
hoped to gain some insight into the educational resourcing issue that can be informative for the
Welsh context. However, most of the evidence that was located comes from the international
literature, particularly the US. Some of the important considerations with respect to
transferability to the Welsh context are listed below:
The policy context. In Wales, funding for compulsory education (up to 16 years of age)
is under the Local Education Authority‘s (LEA) control. In other countries, funding is
partly at the state level and partly at the district level. The source of funds might
influence how decisions are made about the expenditure of available resources. In
Wales (and the UK), for example, ―closing the gap‖ in terms of attainment between low
and average high school leavers is a high policy priority, and so resources could be
geared towards this.
The geographic context. Wales has a substantial number of rural (sparsely populated)
areas, which poses challenges to educational resourcing. These could be increased
costs due to the lack of economies of scale, or through the higher costs for school
transport. Thus, whether rural locations were included in the study was recorded during
data extraction.
The social context. Each country differs in its racial, ethnic, and socioeconomic make-
up. Given the established strong relation between attainment and socioeconomic status
(in particular), it is important to consider these background variables. Where possible,
we note whether such variables were included in the analytical models of the included
studies, which helps to determine how much of the change from the inputs improves the
outcomes above and beyond the background variables.
Matrix Evidence | 20 January 2011 39
The research context. As mentioned above, the data that are used in educational
resourcing studies tend to come from national and large-scale datasets collected by the
educational administration rather than the researchers themselves. In the US, this
typically means data aggregated to the district or even state-level, since the country has
such a vast population. However, the equivalent to district-level in Wales is the LEA.
Using LEA-level data would lose a lot of meaningful variation as Wales has only 22
LEAs. As such, Welsh and UK studies tend to analyse pupil-level data. Thus, there is a
mismatch between the level of analysis between Welsh/UK and US studies (see earlier
discussion about the problems with analysing data at higher levels; see also Vignoles et
al., 2000). We note which data level were used for each study.
Particularly important to the Welsh context, given the devolved educational system in place, is
the issue of autonomy of budget. Fuchs and Wößmann (2007) concluded, ―Student
performance is higher with external exams and budget formulation, but also with school
autonomy in textbook choice, hiring teachers and within-school budget allocations‖ (p. 433).
This supports the system currently in place in Wales. However, other studies noted that simply
giving funding to schools without providing guidance on effective spending practices could lead
to wastage or misspending (e.g., Leuven et al., 2007). Although leaving the decisions in the
hands of the schools is appropriate, it could be beneficial to offer advice and information on
established effective practices that the funding can be spent on, such as reading interventions.
Good uses of funding could be identified trough further evidence reviews.
3.6 Knowledge gaps
Establishing causality and direction of effects
Experimental methods are the strongest designs for evaluating the causal effect of an input on
an outcome, yet very few of the studies in this review employed experimental research
methods. This is likely due to the ethical and social implications of such methods in educational
research—it is difficult to offer increased funding to one school and not another unless one is
substantially deprived in one way to begin with. As a result, most of the studies examined
associations between resourcing and attainment over time (i.e., longitudinal modelling). Such
methods make it difficult to claim confidently any causal relationship between inputs and
outcomes because longitudinal research designs cannot fully take into account initial
differences between individuals, or other variables that might influence the relation between
input and outcome. As such, the findings presented herein generally do not verify a causal
relation, but rather suggest that there may be an association between these variables over time
that appear to have some causal ordering (that is, prior inputs can predict future outcomes).
Hand-in-hand with this issue, it is important to note that many of the studies implicitly adopt a
unidirectional approach, with the assumption that inputs lead to outcomes. However, as Figure
1 suggests, it is possible that outcomes will drive the political agenda (or funding formula) in
such a way that outcomes lead to changes in the inputs. Consider, for example, the case where
low achieving schools are given financial boosters based on their low performance on
Matrix Evidence | 20 January 2011 40
standardised tests. It is likely that a reciprocal effects model, in which resources affect
subsequent outcomes, which in turn affect future resource allocations, is likely to exist (see the
causal loop in Figure 1). This has been termed two-way causality (Levačić, 2007).
This two-way causality can be modelled in longitudinal studies where measures of inputs and
outputs at three or more time points are taken, and analysed using longitudinal structural
equation models. Recent developments in software such as MPlus now allow modelling of
multilevel structure data with structural equation models. This is an exciting new approach to
analysing such data, and should be adopted in future longitudinal studies on this issue.
Pupil-level data
Most of the studies in the review included data that had been aggregated (averaged) to the
school or district level, rather than analysing individual student data. This can lead to errors in
the analytical modelling (see technical discussion in Appendix C). More research should be
conducted using individual student-level data for variables such as attainment in order to gain
more accurate estimates of the relationships.
Teacher characteristics
In the educational resourcing literature, the number of years of teaching experience, the
qualifications of the teachers and teacher salaries are the most commonly measured teacher
characteristics. We argue that teaching experience and the qualification of the teachers are not
necessarily linked to financial resources—rather, they are indicators of the quality of education.
This issue is beyond the scope of the current review and so we do not review these variables
here. However, it is inevitable that the quality of the teachers, their morale, and their
relationships with their students will affect the learning experience. A separate review on the
relationship between resources, quality of teaching, and student outcomes could disentangle
this issue.
Educational interventions
As with teacher characteristics, it is likely that what actually happens in the classroom affects
student outcomes. This in turn is linked to characteristics of the school and its students. For
example, a school with student behavioural problems that spends its money on developing
disciplinary procedures could see more benefits than if it had spent its funding on new library
books. Thus, the effectiveness of educational interventions could be determined by the
characteristics of the sample.
A related consideration is the choice of educational intervention. In our review, we came across
many studies looking at the effectiveness of various programmes, such as improved curriculum
or reading interventions. These were a very heterogeneous set of studies, and beyond the
scope of this review. However, such studies could be reviewed to provide guidance to schools
on possible effective uses of the funds that they received.
Matrix Evidence | 20 January 2011 41
Matrix Evidence | 20 January 2011 42
3.7 References included in the review
Addonizio, M. F., Mullin, C. M., Brown, K. S., Verstegen, D. A., Driscoll, L. G., Alm, J., et al.
(2009). X-Efficiency and Effective Schools: A New Look at Old Theories. Journal of
Education Finance, 35(1), 1–25.
AEL Inc. (2000). Notes from the field: KERA in the classroom. Notes from the field: Education
Reform in Rural Kentucky, 7(1), 1-23.
Anderson, A., Thomas, D. R., Moore, D. W., & Kool, B. (2008). Improvements in school climate
associated with enhanced health and welfare services for students. Learning
Environments Research, 11(3), 245–256.
Arum, R., & LaFree, G. (2008). Educational Attainment, Teacher-Student Ratios, and the Risk
of Adult Incarceration Among US Birth Cohorts Since 1910. Sociology of Education,
81(4), 397–421.
Babcock, P., & Betts, J. R. (2009). Reduced-class distinctions: Effort, ability, and the education
production function. Journal of Urban Economics, 65(3), 314–322.
Bacolod, M., DiNardo, J. E., & Jacobson, M. (2009). Beyond Incentives: Do Schools use
Accountability Rewards Productively? NBER Working Paper.
Baskett, R., Bryant, K., White, W., & Rhoads, K. (2005). Half-day to full-day kindergarten: an
analysis of educational change scores and demonstration of an educational research
collaboration. Early Child Development and Care, 175(5), 419–430.
Betts, J., & Roemer, J. E. (2005). Equalizing opportunity for racial and socioeconomic groups in
the United States through educational finance reform. Department of Economics,
UCSD.
Betts, J. R. (2001). The impact of school resources on women's earnings and educational
attainment: Findings from the National Longitudinal Survey of Young Women. Journal of
Labor Economics, 19(3), 635–657.
Betts, J. R., & Morell, D. (1999). The determinants of undergraduate grade point average: The
relative importance of family background, high school resources, and peer group
effects. The Journal of Human Resources, 34(2), 268–293.
Betts, J. R., Zau, A., & Rice, L. (2003). Determinants of student achievement: New evidence
from San Diego. Public Policy Institute of CA.
Blatchford, P., Bassett, P., Browne, P., Marin, C., & Russell, A. (2004). The effect of class size
on attainment and classroom process in English primary schools (Years 4 - 6) 2000-
2003. DfES Research Report, RRx13 -04.
Blatchford, P., Goldstein, H., Martin, C., & Browne, W. (2002). A study of the class size effects
in English school reception year classes. British Educational Research Journal, 28, 169-
185.
Blatchford, P., Russell, A., Bassett, P., Brown, P., & Martin, C. (2007). The role and effects of
teaching assistants in English primary schools (Years 4 to 6) 2000-2003. Results from
the Class Size and Pupil-Adult Ratios (CSPAR) KS2 Project. British Educational
Research Journal, 33(1), 5–26.
Bradley, S., & Taylor, J. (2002). The effect of the quasi-market on the efficiency-equity trade-off
in the secondary school sector. Bulletin of Economic Research, 54(3), 295–314.
Matrix Evidence | 20 January 2011 43
Brown, J. (2004). Integrated community schools: At the heart of the community. Children Now,
(23 November).
Cramer, K. (2006). Making schools smaller: Do smaller learning communities improve student
outcomes? (PhD thesis). The George Washington University.
Darling-Hammond, L. (2004). The color line in American education: Race, resources, and
student achievement. Du Bois Review: Social Science Research on Race, 1(02), 213–
246.
Dustmann, C., Rajah, N., & Van Soest, A. (2003). Class size, education, and wages. The
Economic Journal, 113(485), 99–120.
Estyn. (2008). The impact of RAISE. Evaluation of the impact of RAISE funding on schools'
effectiveness tackling the link between socio-economic disadvantage and
underachievement. An interim report after the first 18 months. Cardiff: Estyn.
Fuchs, T., & Wössmann, L. (2007). What accounts for international differences in student
performance? A re-examination using PISA data. Empirical Economics, 32(2), 433–464.
Graddy, K., & Stevens, M. (2005). Impact of School Resources on Student Performance: A
Study of Private Schools in the United Kingdom, The. Industrial and Labor Relations
Review, 58(3), 435-431.
Greene, G. K., Huerta, L. A., & Richards, C. (2007). Getting real: A different perspective on the
relationship between school resources and student outcomes. Journal of Education
Finance, 33(1), 49.
Grubb, W. N. (2008). Multiple Resources, Multiple Outcomes: Testing the "Improved" School
Finance with NELS88. American Educational Research Journal, 45(1), 104.
Haimson, L. (2000). Smaller is Better: First-hand reports of early grade class size reduction in
New York City Public Schools. Educational Priorities Panel.
Holmlund, H., McNally, S., & Viarengo, M. (2008). Impact of School Resources on Attainment at
Key Stage 2, Research Report DCSF-RR043.
Hoxby, C. M. (2004). Productivity in Education: The Quintessential Upstream Industry. Southern
Economic Journal, 71(2), 208–232.
Iacovou, M. (2002). Class size in the early years: is smaller really better? Education Economics,
10, 261-290.
Ilon, L., & Normore, A. H. (2006). Relative Cost-Effectiveness of School Resources in Improving
Achievement. Journal of Education Finance, 31(3), 17.
Jacob, B. A., Courant, P. N., & Ludwig, J. (2003). Getting inside Accountability: Lessons from
Chicago. Brookings-Wharton Papers on Urban Affairs, 41–81.
Jenkins, A., Levačić, R., & Vignoles, A. (2006). Estimating the relationship between school
resources and pupil attainment at GCSE (No. RR727). London: Department for
Education and Skills.
Jones, J. T., Toma, E. F., & Zimmer, R. W. (2008). School attendance and district and school
size. Economics of Education Review, 27(2), 140–148.
Kinnucan, H., Zheng, Y., & Brehmer, G. (2006). State Aid and Student Performance: A
SupplyDemand Analysis. Education Economics, 14(4), 487–509.
Koshal, R. K., Koshal, M., & Gupta, A. K. (2004). Students Academic Performance: An
Interaction of Inputs from the Students, Schools, and Voters. Perspectives on Global
Development and Technology, 3(3), 375–394.
Matrix Evidence | 20 January 2011 44
Leuven, E., Lindahl, M., Oosterbeek, H., & Webbink, D. (2007). The effect of extra funding for
disadvantaged pupils on achievement. The Review of Economics and Statistics, 89(4),
721–736.
Levačić, R., & Hardman, J. (1999). The performance of grant maintained schools in England: an
experiment in autonomy. Journal of Education Policy, 14(2), 185–212.
Levačić, R., Jenkins, A., Vignoles, A., Steele, F., & Allen, R. (2005). Estimating the relationship
between school resources and pupil attainment at Key Stage 3. London: Department for
Education and Skills.
Levačić, R., & Marsh, A. (2007). Secondary modern schools: are their pupils disadvantaged?
British Educational Research Journal, 33(2), 155–178.
Loubert, L. (2008). Increasing Finance, Improving Schools. The Review of Black Political
Economy, 35(1), 31–41.
Machin, S. J., McNally, S., Silva, O., Street, H., & Street, G. (2006). New technology in schools:
is there a payoff? (No. IZA DP No. 2234). Discussion Paper Series. Bonn: The Institute
for the Study of Labor (IZA).
McNally, S. (2005). Economic evaluation of the Pupil Learning Credits pilot scheme (No.
Research Brief RB696). London: Department for Education and Skills.
McVicar, D. (1999). School quality and staying-on: resources, peer groups or ethos?. Northern
Ireland Economic Research Centre.
Michie, J. S., Chaney, B. W., & Rockville, M. (2009). Second Evaluation of the Improving
Literacy Through School Libraries Program. No Child Left Behind. Jessup, MD: U.S.
Department of Education Office of Planning, Evaluation and Policy Development.
Miller-Whitehead, M. (2003). Compilation of Class Size Findings: Grade Level, School, and
District. Presented at the Annual Meeting of the Mid-South Educational Research
Association, Biloxi, MS.
Muijs, D. (2003). The effectiveness of the use of learning support assistants in improving the
mathematics achievement of low achieving pupils in primary school. Educational
Research, 45(3), 219–230.
Noden, P., & Schagen, I. (2006). The specialist schools programme: golden goose or conjuring
trick? Oxford Review of Education, 32(4), 431–448.
Parcel, T. L., & Dufur, M. J. (2001). Capital at home and at school: Effects on child social
adjustment. Journal of Marriage and the Family, 63(1), 32–47.
Payne, K. J., & Biddle, B. J. (1999). Poor school funding, child poverty, and mathematics
achievement. Educational Researcher, 28(6), 4.
PricewaterhouseCoopers. (2003). Building Better Performance: an Empirical Assessment of the
Learning and other Impacts of Schools Capital Investment. London: Department for
Education and Skills.
Pugh, G., Mangan, J., & Gray, J. (2008). Resources and attainment at key stage 4: estimates
from a dynamic methodology (DCSF-RB056). Research Brief. Department for Children,
Schools and Families.
Ram, R. (2004). School expenditures and student achievement: evidence for the United States.
Education Economics, 12(2), 169–176.
Reynolds, A. J., Temple, J. A., Robertson, D. L., & Mann, E. A. (2002). Age 21 cost-benefit
analysis of the title I Chicago Child-Parent Centers. Educational Evaluation and Policy
Analysis, 24(4), 267.
Matrix Evidence | 20 January 2011 45
Richmond, P. N. (2007). Wealth and achievement gaps: An examination of Virginia middle
schools (PhD thesis). Old Dominion University.
Steele, F., Vignoles, A., & Jenkins, A. (2007). The effect of school resources on pupil
attainment: a multilevel simultaneous equation modelling approach. Journal of the
Royal Statistical Society-Series A, 170(3), 801–824.
Tikly, L., Osler, A., Hill, J., Vincent, K., Andrews, P., Jeffreys, J., et al. (2002). Ethnic minority
achievement grant: analysis of LEA action plans. Department for Education and Skills.
Van der Klaauw, W. (2008). Breaking the link between poverty and low student achievement:
An evaluation of Title I. Journal of Econometrics, 142(2), 731–756.
Waldfogel, J., & Zhai, F. (2008). Effects of public preschool expenditures on the test scores of
fourth-graders: Evidence from TIMSS. Educational Research and Evaluation, 14(1), 9-
28.
West, A., Pennell, H., Travers, T., & West, R. (2001). Financing school-based education in
England: poverty, examination results, and expenditure. Environment and Planning C,
19(3), 461–472.
Wilson, V., Schlapp, U., & Davidson, J. (2002). Classroom Assistants: Key Issues from the
National Evaluation (No. 1). Insight. Edinburgh: Scottish Executive Education
Department.
Wössmann, L. (2001). Schooling resources, educational institutions, and student performance:
the international evidence. Kiel: Kiel Institute of World Economics.
Yeh, S. S., & Ritter, J. (2009). The Cost-Effectiveness of Replacing the Bottom Quartile of
Novice Teachers Through Value-Added Teacher Assessment. Journal of Education
Finance, 34(4), 426-451.
Matrix Evidence | 20 January 2011 46
Appendix A. Search Strategy, inclusion criteria, and inclusion
figures
A.1 Search sources
The following electronic databases were searched:
Applied Social Sciences Index and Abstracts on the Web (ASSIA). ASSIA is an
indexing and abstracting tool covering health, social services, psychology, sociology,
economics, politics, race relations, and education. It indexes over 500 journals
published in 16 different countries, including the UK and US.
British Education Index (BEI). The BEI provides comprehensive information on
educational research, policy and practice in the UK. It includes journal papers, internet
documents (including Education-line and other sources), conference proceedings, and
British doctoral theses.
EconLit. Published by the American Economic Association, EconLit covers more than
thirty years of economics literature from around the world. It indexes and holds
abstracts for journal articles, books, book reviews, collective volume articles, working
papers and dissertations.
Education Resources Information Center (ERIC). ERIC is the is the world's largest
digital library of education and education-related literature, and holds abstracts and full-
text records of journal articles, books, research syntheses, conference papers, technical
reports, policy papers and other education-related materials.
International Bibliography of the Social Sciences (IBSS). From 1951 onwards. The
essential online resource for social science and interdisciplinary research. Broad
coverage of international journals and books, incorporating over 100 languages and
countries. Abstracts are provided for half of all current journal articles and full text
availability is continually increasing.
Planex. This database holds abstracts of journal articles, books, and government
literature as well as documents such as policies, plans, strategies, and annual reports,
in the following subject areas: urban & rural generation, planning & environment,
transport, social work services, housing, education & lifelong learning, local economic
development.
Web of Knowledge. Includes the Web of Science resource, which is a multidisciplinary
collection of the world‘s leading citation databases, with information from over 10,000
high impact journals and over 110,000 conference proceedings from around the world.
Social Abstracts. CSA Social Services Abstracts provides covers research focused on
social work, human services, and related areas, including social welfare, social policy,
and community development. The database indexes over 1,300 serials publications
since 1979. It holds abstracts of journal articles, dissertations, and citations to book
reviews. As of November 2009, the database holds over 155,000 records.
SPP. Social Policy and Practice is a bibliographic database covering evidence-based
social policy, public health, social services, and mental and community health. It brings
together data from 5 databases: Planex; Acompline; Social Care Online; AgeInfo. The
Matrix Evidence | 20 January 2011 47
database holds around 200,000 records, mainly from the UK, but also from the rest of
Europe and the US. About half the references are to grey literature, including semi-
published reports, working papers, local and central government reports, and material
from the voluntary sector and charities.
Several experts were contacted directly and asked to suggest resources. The following
responded with their suggestions:
Professor Rosalind Levačić, Institute of Education, University of London. As Emeritus
Professor of Economics and Finance of Education, Rosalind is a world leader in this
field. She has worked extensively on the issue of the economics of education, including
working as principal investigator on projects such as ―The Impact of Formula Funding
on Schools‖.
Professor Dave Adamson, OBE, University of Glamorgan. Professor of Community &
Social Policy and Director of the University of Glamorgan‘s Centre for Regeneration and
Sustainable Communities, Director of the Programme for Community Regeneration and
Chair of the Regeneration Hub, an inter-disciplinary group at Glamorgan which brings
together a number of research Centres and Units engaged with the wider regeneration
agenda. Working in this field he has had three periods as an Advisor to the Welsh
Assembly Government on issues of regeneration and worked on the design of the
‗Communities First‘ programme.
Professor Stephen Gorard, University of Birmingham. Committed to the improvement of
education in terms of effectiveness and equity, his research is ‗society-wide‘ and lifelong
in scope. He has conducted studies of primary education, early childhood, secondary
education, FE, HE, adult and continuing education, and informal learning in the home.
Professor Fiona Steele, University of Bristol. A Professor of Social Statistics, Fiona has
published on assessing the impact of school resources on pupil attainment.
A.2 Search strategy
The full record (title, abstract and keywords) was searched in all databases. The long search
strategy (for those databases that do not have strict character limitations, such as ERIC) was as
follows:
(schools or educat*) AND (funding or resourc* or invest* or cost* or
class size or physical environment* or structur* or architecture* or
teacher pupil ratio* or teacher student ratio* or classroom help* or
classroom assist* or teaching assist* or learn* assist* or book* or
instruction time or IT equipment* or Sport* equipment* or music*
equipment* or “Revenue Support Grant” or “Local Government funding” or
entitlement*) AND (attainment* or achievement* or “minimum standard”
or “league table” or test* score or exam* or social disadvant* or
“drop-out” or participation or progression or transition or social
inclusion* or social exclusion* or social isolation* or social
deprivation* or material deprivation* or inequality* or household
income* or poverty or free school meal*)) OR (Wales or Welsh)
Matrix Evidence | 20 January 2011 48
The short search strategy (for those databases that do have strict character limitations, such as
BEI) was as follows:
(schools or educat*) AND (funding or resource or grant$ or invest$ or
cost$ or (ratio and (teacher pupil or teacher student)) or classroom
help$ or (assist$ and (classroom or teaching or learning)) or
(equipment$ and (Sport$ or music or IT)) or entitlement$) AND (attain$
or achieve$ or minimum standards or league tables or exam$ or
participation or transition or (social and (inclusion or exclusion or
isolation or disadvantage or deprivation )) or inequality or poverty)
OR (Wales or Welsh)
The full record (title, abstract and keywords) was searched in all databases. A date restriction
―1999 – Current‖ was used to maximise the relevancy of the evidence. Searches were not
restricted by country or language, although only English and Welsh language abstracts were
retrieved.
A.3 Screening
The inclusion criteria used in the screening stage were the following:
1. The study is available in English or Welsh language
2. The study refers to the use or allocation of resources or funding
3. The study refers to educational settings—specifically, primary and/or secondary schools
(i.e., educational institutions covering students that were within the age range 4-16
years at the time of attending). Note that this does not mean that the participants are
limited to those age ranges, but rather that the resource allocation referred to in (2) had
to occur while the students were in schools for 4-16 year olds. Thus, any resource
allocation to universities, for example, would fail this criterion.
4. The study includes the measurement of outcomes such as attainment, drop out, school
exclusion, post school participation in education/training
5. The study provides quantitative or qualitative data on the outcomes mentioned in (4), or
is a systematic review (or meta-analysis) of the relevant literature. Non-systematic
literature reviews, ‗think pieces‘, or policy pieces should not be included.
6. The study was published in 1999 or after
7. The study was carried out in a context which ensures some level of transferability.
Those studies that, for example, are concerned with expanding the coverage of
compulsory education should be excluded.
Matrix Evidence | 20 January 2011 49
Table A1.
Inclusion figures by search source (including 3 review studies)
Abstract Full text Total
Ex NA EX1 EX2 EX3 EX4 EX5 EX6 IN Irretriev Dup Linked EX1 EX2 EX3 EX4 EX5 EX6 EX7 IN
ASSIA 0 0 32 11 11 0 0 6 0 0 0 0 4 0 1 0 0 0 1 60
BEI 489 1 191 29 62 1 29 44 6 1 0 0 20 0 7 1 0 1 8 846
EconLit 3 0 122 43 82 2 1 47 5 0 3 13 1 5 3 3 1 13 300
ERIC 0 0 54 4 169 7 0 61 14 0 0 0 20 0 4 14 0 2 7 295
IBSS 15 0 130 36 54 2 0 31 1 4 1 1 11 2 1 0 0 1 9 268
Planex 0 0 10 63 17 0 0 9 2 1 0 0 3 1 1 1 0 0 0 99
Soc Abs 0 0 536 65 70 7 0 24 5 1 0 0 13 0 0 0 0 1 4 702
SPP 3 0 250 110 283 3 0 30 1 0 2 0 5 1 5 3 0 0 13 679
Wok 0 9 271 86 82 9 0 30 4 1 0 0 9 0 6 5 0 2 3 487
Experts 0 0 0 0 0 0 0 12 0 1 0 0 8 0 0 1 1 0 5 12
Total 510 10 1596 447 830 31 30 298 38 9 6 1 106 5 30 28 4 8 63 3752
Matrix Evidence | 20 January 2011 50
Appendix B. Flow of literature through the review
Figure B. Flow of literature
Excluded on abstract n = 510 EX 1 n = 10 EX 2 n = 1,596 EX 3 n = 447 EX 4 n = 830 EX 5 n = 31 EX 6 n = 30
References located: database searches n = 3,736 provided by experts n = 16
Excluded on abstract n = 3,454
Excluded on full text EX 1 n = 1 EX 2 n = 106 EX 3 n = 5 EX 4 n = 30 EX 5 n = 28 EX 6 n = 4 EX 7 n = 8
Included studies n = 60 + 3 reviews
Irretrievable n = 38
Full text retrieval n = 298
Excluded on full text n = 182
Duplicate and Linked studies n = 15
Matrix Evidence | 20 January 2011 51
Appendix C. Study characteristics of the primary studies
C.1 Descriptive statistics
Table C1.
Study publication year (primary studies)
Year of Study n
1999 4
2000 2
2001 5
2002 5
2003 6
2004 7
2005 4
2006 7
2007 7
2008 10
2009 3
TOTAL 60
Table C2
Country of study (primary studies)
Country n
England 17
Wales 1
England and Wales
3
Northern Ireland 1
Scotland 2
UK 1
US 30
Other 2
Multi-country 3
Total 60
Matrix Evidence | 20 January 2011 52
C.3 Limitations with the individual studies included in this review (an
annotated bibliography)
Addonizio (2009). The dataset used had a large number of missing data points as a result of
non-participation by schools in the study; schools that did not participate may be
systematically different from those that did, which could bias the results. Also, the study
used school-level data, and did not control for many other variables. This makes the
findings from these analyses questionable.
AEL Inc (2000). The authors noted that there were some problems with introducing aspects of
the reform, suggesting the changes in resources may not have been consistent across
the sample. Moreover, many initiatives were introduced at once, and the evaluation was
not rigorous. It is therefore hard to tell how much of the improvements were due to
particular changes.
Anderson et al. (2008). The reliability of some of the scales used to measure school climate
were not as high as desired. This could lead to measurement error of the instrument.
Perhaps more concerning, the time between the 2003 and 2004 student surveys was
only 5 months, in which the effects of the programme may not have had time to manifest.
This could explain the non-significant findings for this analysis.
Arum and LaFree (2008). The authors note that a limitation of their research lies in linking PTRs
and incarceration risk. ―In particular, our data do not allow us to determine the extent to
which teacher/student ratios affect incarceration risk by improving youth socialization and
decreasing criminality or through their impact on the decision-making of officials in the
legal system‖ (p. 26).
Babcock and Betts (2009). Used gain scores to account for prior attainment. They noted that
that there were larger attainment gains for disadvantaged students.
Bacolod et al. (2007). A conclusion of this study was that the reward money given to schools
was not used in meaningful way, undermining the resource allocation. Moreover, the
study treats those that did not receive the reward as a control group. This is dubious as
there may be initial differences between the groups.
Baskett et al. (2004). It was not clear how many schools were involved in the project, or the
reliability of the measures. The groups analysed were not really comparable.
Betts (2001). This study uses very old data: the school expenditure data were collected in 1968,
while data on attainment and wages was collected between 1968 and 1991. It is difficult
to tell whether this still has enough currency.
Betts et al. (2003). Data consisted of gain scores rather than raw attainment scores, which
helps to account for prior attainment.
Matrix Evidence | 20 January 2011 53
Betts and Morell (1999). The sampling of the study makes it difficult to generalise to other
contexts; however, the inclusion of social background variables helps to isolate the
variance due to PTR.
Betts and Roemer (2005). There is limited information on the sampling strategy, making it
difficult to determine the representativeness of the sample.
Blatchford et al. (2004). This report does not adequately describe the statistical analyses used
to arrive at the conclusions.
Blatchford et al. (2002). A longitudinal design with baseline assessment was used. They also
ran models with different ―splines‖, because ―Almost all previous studies have limited
themselves to linear relationships between achievement and class size, but this may only
be true over a restricted range. Fitting simple polynomial relationships may impose too-
rigid constraints on the shape of the relationship, especially at the extremes (p. 173)‖.
Blatchford et al. (2007). The study examined relationships between TAs and the academic
outcomes for the whole class, rather than for the specific pupils that they support.
Bradley and Taylor (2002). This study notes that funding reform introduced both financial and
organisational changes, although the extent of these inputs were not assessed.
Brown (2004). There is a strong possibility of confounding variables in this study.
Cramer (2006). This study used a non-random comparison group.
Darling-Hammond (2004). These models take into account a reasonable number of background
variables, which reduces the risk of endogeneity; however, the use of aggregated data
raises concerns about the strength of the observed relationship.
Dustmann et al. (2003). Although the study did control for prior attainment, it was based on
quite old data (collected in 1974 and 1981).
Estyn (2008). The authors note that the diversity of the RAISE funded activities makes the
assessment of effectiveness difficult. ―It is often difficult to isolate one activity or
intervention as the main cause of any improvement in pupil performance. It is also difficult
to assess the impact of many RAISE activities in terms of improved standards because
many have not run for long enough to see clear trends in outcomes‖ (p. 9).
Fuchs and Wößmann (2007) controlled for numerous family-background and institutional
effects, lending confidence to the findings.
Matrix Evidence | 20 January 2011 54
Graddy and Stevens (2005). This study controlled for prior attainment, which means that any
observed differences as a result of school characteristics should be unique to the school
(rather than attributable to differences in the initial abilities of the students).
Greene et al. (2007). The study ignores the multilevel data structure, and presents an unclear
justification for inclusion of "number of feeder schools" in the model.
Grubb (2008). The study attempts to make causal attributions from correlational data.
Haimson (2000). Although it is claimed that the schools are differentiated to reflect a variety of
implementations of the reduced class sizes funding, it is not really explained how those
specific schools were chosen. Perhaps more importantly, the outcome measure is based
on perception of teachers rather than actual measured performance.
Holmlund et al. (2008). This study took into account the multilevel structure of the data. Also,
the authors modelled a range of background variables, including deprivation.
Hoxby (2004). It is unclear how consistent the general increase in expenditure is manifested at
the school-level, which is a function of the aggregation bias issue.
Iacovou (2002) included the interaction between school size and school type ―thus freeing the
estimates of bias arising from the re-distributive allocation of educational resources faced
by the children in this sample (p. 282)‖. Although this paper uses careful longitudinal
models, the data is based on pupils that were born in 1958, and state that class sizes for
this cohort were typically much higher than today‘s classes because of the baby boom at
this time. In fact, around 80% of the students were in classes with more than 30 students.
Thus, what is considered a ―small‖ class in this sample is on a different scale to more
recent studies.
Ilon and Normore (2006) provided no information about the sample to gauge whether the results
of this study are generalisable to other contexts, and used school-level data raising
concerns about aggregation bias. Also, the counterintuitive finding about the negative
relation between expenditure and attainment could be due to the simple regression
analysis used, because other variables could explain the effect (e.g., class size and
teacher experience could have some confounding with expenditure).
Jacob (2003). The estimates produced by the analyses had large standard deviations, which
suggests that there might be differences between schools that are not accounted for in
the analytical modelling. Overall, we can have low confidence in transferring findings from
this study to the Welsh school system.
Jenkins et al. (2006). This study appropriately controlled for endogeneity by including various
control variables, such as prior attainment, school size, and political control of the LEA.
This also minimises the likelihood that results are due to aggregation bias associated with
using school-level data.
Matrix Evidence | 20 January 2011 55
Jones et al. (2006). There are a large number of predictor variables in model—perhaps more
than the amount of data can sufficiently model. They appear to have averaged findings
across 9 years, and the test for time effects does not seem adequate to detect changes.
Kinnucan et al. (2006). This article tests various explanatory models, although it is not clear to
what extent they're related to the actual empirical data.
Klaauw (2008). The study does not adequately report what the funding was spent on.
Koshal et al. (2004) demonstrated the importance of accounting for background and
instrumental variables in educational resourcing studies.
Leuven et al. (2007). The discussion in the paper suggests that seems that money was not
spent on its intended purpose, which could explain the effects. However, no data were
collected to confirm this suspicion. There were also some concerns with the how the
sample was determined, as disadvantage was not clearly defined.
Levačić et al. (2005) based their analyses on a good pupil-level database, which increases
confidence in the findings.
Levačić and Hardman (1999). This study used carefully selected instrumental variables
(including student background and school governance variables) in the modelling, which
helps to distinguish how much of the variance in attainment across schools is actually
due to the different funding mechanisms. However, this study did not have pupil level
data and did not sufficiently control for pupil background effects.
Levačić and Marsh (2007). Analysis could not be conducted that brought budgetary variables
and attainment variables into the same model, and so it is difficult to determine whether
this relation was significant or not.
Loubert (2008) used data from 1990 and 1997 for an interrupted time series analysis design,
which they acknowledged was a limitation given the dramatic change in the racial make-
up of the population during that time.
Machin et al. (2006). This study uses good representative sampling and well-developed
datasets.
McNally (2005). This has an odd design with 2 control groups, one of which seems to include
the other (see p. 7-8).
McVicar (1999). This study posits causal relations although it used a cross-sectional survey
design.
Matrix Evidence | 20 January 2011 56
Michie (2009). The data are insufficient, with lots of missing data—only 40% of the schools
completed the baseline attainment test.
Miller-Whitehead (2003). Being a conference paper, this study lacks a lot of detail for fully
interpreting the findings.
Noden and Schagen (2006). As with many of the studies in this review, establishing a link
between funding and attainment does not determine whether improvements are due to
the characteristics of the programme or to the additional funding provided.
Parcel and Dufur (2001). This uses a strong dataset and models both school and family
variables.
Payne and Biddle (1999). In addition to concerns with using aggregated data, there is a problem
with the gap between the time that student attainment data was gathered (1982) and the
school/district information was collected (1988-1990). It is not possible to know whether
the school conditions observed in 1988-1990 are equivalent to those experienced when
the student attainment data were collected. This makes any interpretation of the findings
dubious.
PricewaterhouseCoopers (2003). Difficult to disentangle funding effects from the effect of actual
initiatives implemented as a result of the extra capital.
Pugh et al. (2008). The estimates pertaining to some school types are ―subject to a greater
margin of error‖ (p. 2).
Ram (2004). The use of the state-level average expenditure might mask the intrastate
(between-school and between-student) variance in spending and student attainment.
Reynolds et al. (2001). The design used a non-random control group comparison (i.e., quasi-
experimental design). Some of the estimates were based on projected rather than
measured effects.
Richmond (2008). The data do not enable causal conclusions to be drawn.
Steele et al. (2007) incorporated a number of instrumental variables, but the datasets they used
did not allow them to model the effects of ‗home‘ variables. The authors acknowledge that
such variables might have been useful in explaining the odd negative effect of increased
funding on English attainment.
Tikly et al. (2002). There were some serious gaps in the data.
Waldfogel and Zhai (2008). This study is strengthened by the use of a strong international
database; however, the authors note that there may be factors across the countries that
they did not control for that might explain differences in student attainment.
Matrix Evidence | 20 January 2011 57
West et al. (2001). In addition to concerns about using district-level aggregated data, the nature
of the data do not allow causal attributions to be made.
Wößmann’s (2001) findings were based on cross-sectional data, which does not allow
prediction of prior effects on future outcomes.
Matrix Evidence | 20 January 2011 58
Appendix D. Data extraction tool
Study information
1. Author/s & year. Use the format: Surname (year). E.g., Matrix & King (2009).
2. Type of publication. E.g. Journal article, government report, etc.
3. Funder of research.
4. Author(s) organizations(s).
5. Country of study
6. Specific location
7. Urban/rural. Choose one if specified.
8. Wales specific data? Y/N
Methodology and quality assessment
9. Aims of study. Stated objectives of the study.
10. Type of school. (e.g., primary, secondary, public, private)
11. Number of schools. How many schools were measured in total?
12. Number of students/participants. How many people were measured in total?
13. Special sample characteristics. (e.g. low-achieving students; focus on an ethic
grup)
14. Sampling method. Note any details about the sampling method (e.g., random
sampling, purposive sampling)
15. Sampling process described? Y/N. Does the study adequately describe the
sampling process, sufficient for replication?
16. Sample: socio-demographic data? Y/N. Is there a description of background
variables such as age, gender, race, or SES for the sample as a whole? Select ―Y‖
if any are present.
17. What was the study design as stated in the paper? E.g., RCT, quasi-experiment,
observational, literature review, secondary data analysis, etc. If it is not explicitly
mentioned, then infer from the methods section.
18. What is the main data type?
a. Quantitative
b. Qualitative
c. Mixed
d. Other (please note)
19. How were data collected?
e. Survey (closed and/or open questions)
f. Interviews
g. Observation
h. Existing data sets
i. Literature searching
j. Case study
k. Focus groups
l. Other (please note)
20. Source of data (e.g., name and date of survey, interviewees)
Matrix Evidence | 20 January 2011 59
21. Data collection described? Y/N. Is there a description of the data collection
process, sufficient for replication?
Findings
22. Type of resource allocation (e.g., money, facilities, etc)
23. Source of resource allocation (e.g., local govt, central govt, etc)
24. Attainment outcomes. Note the results related to attainment or attainment.
25. Other outcomes. List all other (non-attainment) relevant outcomes and their
results.
26. Possible data for meta-analysis? Y/N. Use ―Y‖ if there appears to be sufficient data
to calculate effect sizes for the main outcomes. Note the page number where the
data can be found.
27. Key findings as stated by the authors.
28. Limitations of the study, as stated by the authors.
Other
29. Link to full text if available
30. Notes. Any other questions of interest. E.g. note external validity issues, or any
other QA-related information.
31. Coder. Who coded the record?
Matrix Evidence | 20 January 2011 60
Appendix E. Summary of included studies
Table E overleaf summarises the resources and outcome variables reported in the included
studies.
Matrix Evidence | 20 January 2011 61
Table E Summary of included studies (primary studies only)
Resources Outcomes
Citation Country Type of school
Per pupil expend-iture
Class size, student teacher ratio
Teacher characteristics
Teaching assistants
Facilities/ books
Multi-faceted grant/ money
Attain-ment
Education progression/ participation
Truancy/ attendance
Wages/ employ-ment Other
Addonizio (2009) USA primary
AEL Inc (2000) USA primary Anderson et al. (2008)
New Zealand secondary
Arum & LaFree (2008) USA secondary Babcock & Betts (2009) USA primary
Bacolod et al. (2007) USA
primary, middle and secondary
Baskett et al. (2004) USA kindergarten
Betts & Morell (1999) USA secondary Betts & Roemer (2005) USA secondary
Betts (2001) USA secondary
Betts et al. (2003) USA primary, middle, and secondary
Blatchford et al (2002) England Infant, primary Blatchford et al (2004) England primary Blatchford et al (2007) England primary Bradley & Taylor (2002) England secondary
Brown (2004) Scotland secondary
Cramer (2006) USA secondary Darling-Hammond (2004) USA school district
Dustmann et al. (2003)
England and Wales secondary
Matrix Evidence | 20 January 2011 62
Resources Outcomes
Citation Country Type of school
Per pupil expend-iture
Class size, student teacher ratio
Teacher characteristics
Teaching assistants
Facilities/ books
Multi-faceted grant/ money
Attain-ment
Education progression/ participation
Truancy/ attendance
Wages/ employ-ment Other
Estyn (2008) Wales
primary, secondary, special education
Fuchs & Wößmann (2007)
31 countries secondary
Graddy & Stevens (2005) UK secondary Greene et al. (2007) USA secondary
Grubb (2008) USA secondary
Haimson (2000) USA primary Holmlund et al. (2008) England Primary
Hoxby (2004) USA primary and secondary
Iacovou (2002) UK Infant, primary Ilon & Normore (2006) USA primary
Jacob (2003) USA primary Jenkins et al. (2006) England secondary Jones et al. (2006) USA secondary Kinnucan et al. (2006) USA middle
Klaauw (2008) USA primary and middle
Koshal et al. (2004) USA secondary
Leuven et al. (2007)
the Netherlands primary
Levačić & Hardman (1999)
England and Wales
primary and secondary
Levačić & Marsh (2007) England secondary
Matrix Evidence | 20 January 2011 63
Resources Outcomes
Citation Country Type of school
Per pupil expend-iture
Class size, student teacher ratio
Teacher characteristics
Teaching assistants
Facilities/ books
Multi-faceted grant/ money
Attain-ment
Education progression/ participation
Truancy/ attendance
Wages/ employ-ment Other
Levačić et al. (2005) England secondary
Loubert (2008) USA primary
Machin et al. (2006) England
primary and secondary
McNally (2005) England secondary
McVicar (1999) Northern Ireland secondary
Michie (2009) USA primary and secondary
Miller-Whitehead (2003) USA primary Muijs & Reynolds (2003) England primary Noden & Schagen (2006) England secondary Parcel & Dufur (2001) USA primary Payne & Biddle (1999) USA secondary
PricewaterhouseCoopers (2003) England
primary and secondary
Pugh et al. (2008) England secondary
Ram (2004) USA primary and secondary
Reynolds et al. (2001) USA primary
Richmond (2008) USA middle Steele et al. (2007) UK secondary
Tikly et al. (2002) England primary Waldfogel & Zhai (2008) 7 countries primary
West et al. (2001) England Local Education Authorities
Wilson et al. (2002) Scotland primary
Matrix Evidence | 20 January 2011 64
Resources Outcomes
Citation Country Type of school
Per pupil expend-iture
Class size, student teacher ratio
Teacher characteristics
Teaching assistants
Facilities/ books
Multi-faceted grant/ money
Attain-ment
Education progression/ participation
Truancy/ attendance
Wages/ employ-ment Other
Wößmann (2001) 39 countries secondary